AI research atlas / v2
Learn AI papers in the right order.
Start with landmark ideas, move through foundations, then branch into LLMs, GenAI, agents, systems, and safety with a reading path that keeps the field from feeling random.
Build the mental timeline before going deep.
Move from foundations to modern systems.
Learning path
Where to start, and what to read next
Orientation / 1-2 weeks
Start Here
Read the papers everyone keeps referencing so the rest of the map has anchors.
Foundations / 2-4 weeks
Classical ML
Learn the statistical and probabilistic ideas that still sit under modern models.
Foundations / 1-2 weeks
Optimization
Understand the training mechanics behind gradient-based learning.
Builder / 3-5 weeks
Deep Learning Core
Move through representation learning, CNNs, residual networks, and scaling patterns.
Builder / 3-6 weeks
Sequence Models and LLMs
Study attention, transformers, language modeling, instruction tuning, and evaluation.
Specialist / 3-6 weeks
Generative AI
Compare GANs, diffusion, autoregressive generation, and modern GenAI workflows.
Specialist / 2-4 weeks
Multimodal and Retrieval
Connect language with images, retrieval, embeddings, and real-world knowledge access.
Specialist / 3-5 weeks
RL and Agents
Learn decision making, feedback, policy learning, and agent-style systems.
Practitioner / 2-4 weeks
Systems and Scaling
Understand the infrastructure and engineering papers behind large-scale training.
Practitioner / 2-4 weeks
Safety and Interpretability
Study robustness, alignment, transparency, and how to reason about model behavior.
Learning Paradigms
Trust and Deployment
Research library
Generative AI
Showing papers for this learning path. Open any paper card to read the full paper and related resources.
MizAR 60 for Mizar 50
As a present to Mizar on its 50th anniversary, we develop an AI/TP system that automatically proves about 60% of the Mizar theorems in the hammer setting. We also automatically prove 75% of the Mizar theorems when the automated provers are helped by using only the premises used in the human-written Mizar proofs. We describe the methods and large-scale experiments leading to these results. This includes in particular the E and Vampire provers, their ENIGMA and Deepire learning modifications, a number of learning-based premise selection methods, and the incremental loop that interleaves growing a corpus of millions of ATP proofs with training increasingly strong AI/TP systems on them. We also present a selection of Mizar problems that were proved automatically.
GAN（Generative Adversarial Nets）
We propose a new framework for estimating generative models via an adversarial process, in which we simultaneously train two models: a generative model G that captures the data distribution, and a discriminative model D that estimates the probability that a sample came from the training data rather than G. The training procedure for G is to maximize the probability of D making a mistake. This framework corresponds to a minimax two-player game. In the space of arbitrary functions G and D, a unique solution exists, with G recovering the training data distribution and D equal to ½ everywhere. In the case where G and D are defined by multilayer perceptrons, the entire system can be trained with backpropagation. There is no need for any Markov chains or unrolled approximate inference networks during either training or generation of samples. Experiments demonstrate the potential of the framework through qualitative and quantitative evaluation of the generated samples.
Auto-Encoding Variational Bayes
How can we perform efficient inference and learning in directed probabilistic models, in the presence of continuous latent variables with intractable posterior distributions, and large datasets? We introduce a stochastic variational inference and learning algorithm that scales to large datasets and, under some mild differentiability conditions, even works in the intractable case. Our contributions are two-fold. First, we show that a reparameterization of the variational lower bound yields a lower bound estimator that can be straightforwardly optimized using standard stochastic gradient methods. Second, we show that for i.i.d. datasets with continuous latent variables per datapoint, posterior inference can be made especially efficient by fitting an approximate inference model (also called a recognition model) to the intractable posterior using the proposed lower bound estimator. Theoretical advantages are reflected in experimental results.
High-Resolution Image Synthesis with Latent Diffusion Models
By decomposing the image formation process into a sequential application of denoising autoencoders, diffusion models (DMs) achieve state-of-the-art synthesis results on image data and beyond. Additionally, their formulation allows for a guiding mechanism to control the image generation process without retraining. However, since these models typically operate directly in pixel space, optimization of powerful DMs often consumes hundreds of GPU days and inference is expensive due to sequential evaluations. To enable DM training on limited computational resources while retaining their quality and flexibility, we apply them in the latent space of powerful pretrained autoencoders. In contrast to previous work, training diffusion models on such a representation allows for the first time to reach a near-optimal point between complexity reduction and detail preservation, greatly boosting visual fidelity. By introducing cross-attention layers into the model architecture, we turn diffusion models into powerful and flexible generators for general conditioning inputs such as text or bounding boxes and high-resolution synthesis becomes possible in a convolutional manner. Our latent diffusion models (LDMs) achieve new state of the art scores for image inpainting and class-conditional image synthesis and highly competitive performance on various tasks, including unconditional image generation, text-to-image synthesis, and super-resolution, while significantly reducing computational requirements compared to pixel-based DMs.
A survey on Image Data Augmentation for Deep Learning
Deep convolutional neural networks have performed remarkably well on many Computer Vision tasks. However, these networks are heavily reliant on big data to avoid overfitting. Overfitting refers to the phenomenon when a network learns a function with very high variance such as to perfectly model the training data. Unfortunately, many application domains do not have access to big data, such as medical image analysis. This survey focuses on Data Augmentation, a data-space solution to the problem of limited data. Data Augmentation encompasses a suite of techniques that enhance the size and quality of training datasets such that better Deep Learning models can be built using them. The image augmentation algorithms discussed in this survey include geometric transformations, color space augmentations, kernel filters, mixing images, random erasing, feature space augmentation, adversarial training, generative adversarial networks, neural style transfer, and meta-learning. The application of augmentation methods based on GANs are heavily covered in this survey. In addition to augmentation techniques, this paper will briefly discuss other characteristics of Data Augmentation such as test-time augmentation, resolution impact, final dataset size, and curriculum learning. This survey will present existing methods for Data Augmentation, promising developments, and meta-level decisions for implementing Data Augmentation. Readers will understand how Data Augmentation can improve the performance of their models and expand limited datasets to take advantage of the capabilities of big data.
Review of deep learning: concepts, CNN architectures, challenges, applications, future directions
In the last few years, the deep learning (DL) computing paradigm has been deemed the Gold Standard in the machine learning (ML) community. Moreover, it has gradually become the most widely used computational approach in the field of ML, thus achieving outstanding results on several complex cognitive tasks, matching or even beating those provided by human performance. One of the benefits of DL is the ability to learn massive amounts of data. The DL field has grown fast in the last few years and it has been extensively used to successfully address a wide range of traditional applications. More importantly, DL has outperformed well-known ML techniques in many domains, e.g., cybersecurity, natural language processing, bioinformatics, robotics and control, and medical information processing, among many others. Despite it has been contributed several works reviewing the State-of-the-Art on DL, all of them only tackled one aspect of the DL, which leads to an overall lack of knowledge about it. Therefore, in this contribution, we propose using a more holistic approach in order to provide a more suitable starting point from which to develop a full understanding of DL. Specifically, this review attempts to provide a more comprehensive survey of the most important aspects of DL and including those enhancements recently added to the field. In particular, this paper outlines the importance of DL, presents the types of DL techniques and networks. It then presents convolutional neural networks (CNNs) which the most utilized DL network type and describes the development of CNNs architectures together with their main features, e.g., starting with the AlexNet network and closing with the High-Resolution network (HR.Net). Finally, we further present the challenges and suggested solutions to help researchers understand the existing research gaps. It is followed by a list of the major DL applications. Computational tools including FPGA, GPU, and CPU are summarized along with a description of their influence on DL. The paper ends with the evolution matrix, benchmark datasets, and summary and conclusion.
Attention Is All You Need
The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data.
The Multimodal Brain Tumor Image Segmentation Benchmark (BRATS)
In this paper we report the set-up and results of the Multimodal Brain Tumor Image Segmentation Benchmark (BRATS) organized in conjunction with the MICCAI 2012 and 2013 conferences. Twenty state-of-the-art tumor segmentation algorithms were applied to a set of 65 multi-contrast MR scans of low- and high-grade glioma patients-manually annotated by up to four raters-and to 65 comparable scans generated using tumor image simulation software. Quantitative evaluations revealed considerable disagreement between the human raters in segmenting various tumor sub-regions (Dice scores in the range 74%-85%), illustrating the difficulty of this task. We found that different algorithms worked best for different sub-regions (reaching performance comparable to human inter-rater variability), but that no single algorithm ranked in the top for all sub-regions simultaneously. Fusing several good algorithms using a hierarchical majority vote yielded segmentations that consistently ranked above all individual algorithms, indicating remaining opportunities for further methodological improvements. The BRATS image data and manual annotations continue to be publicly available through an online evaluation system as an ongoing benchmarking resource.
Denoising Diffusion Probabilistic Models
DiffuCpG 1. Introduction In this study, we used a generative AI diffusion model to address missing methylation data. We trained the model with Whole-Genome Bisulfite Sequencing data from 26 acute myeloid leukemia samples and validated it with Reduced Representation Bisulfite Sequencing data from 93 myelodysplastic syndrome and 13 normal samples. Additional testing included data from the Illumina 450k methylation array and Single-Cell Reduced Representation Bisulfite Sequencing on HepG2 cells. Our model, DiffuCpG, outperformed previous methods by integrating a broader range of genomic features, utilizing both short- and long-range interactions without increasing input complexity. It demonstrated superior accuracy, scalability, and versatility across various tissues, diseases, and technologies, providing predictions in both binary and continuous methylation states. In this repository, we deposit the code used to build the diffusion models along with necessary example datasets to train and test a diffusion model for methylation imputation purposes. Docker Usage Install Docker Install Docker using the following link:https://docs.docker.com/engine/install/Recommended system specs: Debian 12 bookworm with 16GB RAM or more.Make sure you have the latest Nvidia GPU driver installed and docker can access your Nvidia GPU. Run Docker images with Tissue-specific Models docker pull yay135/diffucpg_tssUse our example to generate input samples with Hi-C matrix and CIS (Confidence Interval Cross Sample) data.docker run -it yay135/diffucpg_tssthenpython generate_train_test_samples.py The tissue-specific models (pytorch) are for CD34+ cells, GBM and BRCA, they are stored in folders named "model*" in the image. Run the Tissue specific modelsdocker run -it yay135/diffucpg_tssthenpython batch_run.py Run Docker images Example Models docker pull yay135/diffucpgIf you do not have a GPU enabled system, pull a CPU-only imagedocker pull yay135/diffucpg_cpuprepare your input data directory, use the following command to print a example input data directorydocker run --rm yay135/diffucpg -e trueassume your data directory name is "input_data"in windowsdocker run --gpus all -v .\input_data\:/data --rm yay135/diffucpgin unix or linuxdocker run --gpus all -v ./input_data:/data --rm yay135/diffucpg Other docker options -d or --device : select which cuda device to run with, default is 0-m or --mingcpg : scan your methyl array, limit only imputing windows with at least m non-missing methyl values, default is m=10-o or --overlap : set number of impute epochs, shift window locations between epochs, get mean imputed values for each CpG location, default is 2example:docker run --gpus all -v ./input_data:/data --rm yay135/diffucpg -d 1 -m 5 -o 3use cuda device 1, min number of non-missing methyl values in a window is 5, overlap epochs 3 The following tutorials are for non-docker usages. 2. Data and Models Example datasets are available for download using "gdown.sh". The example datasets only contain WGBS methylation data. The model is the DDPM diffusion model, the repository contains a complete implementation for 1-dimensional input. Please refer to https://arxiv.org/abs/2006.11239 and https://huggingface.co/blog/annotated-diffusion for more details. 3. How to use 3.1 System Requirements The number of steps in the diffusion process is set to 2000. Imputing a sample requires 2000 steps. Gpu acceleration is preferred. 16GB of RAM is required. The code is fully tested and operational on the following platform: Distributor ID: DebianDescription: Debian GNU/Linux 12 (bookworm)Release: 12Codename: bookworm 3.2 Clone the Current Project Run the following command to clone the project.git clone https://github.com/yay135/DiffuCpG.git 3.4 Configure Environment Make sure you have the following software installed in your system:Python 3.9+Pytorch 2.0.1+ 3.4 Run Training and Testing python run.pyThe script will download necessary data and install dependencies automatically. 4 Data and Script Details 4.1 RAW Data The methylation arrays downloaded are in the folder "raw", each file is a methylation array. The first 2 columns are "chromosome" and "location". The assembly used for mapping in our project is the "GRCH37 primary assembly". It is also downloaded automatically. The rest of the columns in each file are methylation levels(required) and other biological data (optional) you wish to incorporate to enhance the model. These files in the raw folder are the initial inputs for pipeline,if you wish to use your own data, it must be configured as such before running the pipeline. 4.2 Generate Sample Use script "generate_samples.py" to generate samples for training and testing.The model can not directly read and impute a methylation array file. Instead, each methylation array is divided into windows, each window is 1kb (1000 base pairs) in length, and each training testing sample is generated from a window. Each sample contains at least 5 channels. the first 4 is the sequence one-hot encoding, the 5th is the methylation data. If a base pair location is not a CpG location, the methylation data value for it is "-1". If a CpG's methylation data is missing or waiting for imputaion, its value is also "-1". Other biological data can be added as extra channels. Check out example raw files in the folder "raw" to form your own datasets for training and testing sample generation.For each raw file in the "raw" folder, the first 3 columns are chr, loc, and methylation.The rest of the columns are treated as additional channels and will be added to each sample during generation. '-d' or '--folder': specify raw data folder'-i' or '--index' : which column in a raw file is the methylation array'-t' or '--tol' : how many missing methylation value is tolerated(we recommend 0 for generating training samples and -1 for generating testing samples, 0 will force the script to only select from windows with no missings, -1 will tolerate missing as much as possible.)'-c' or '--chr' : limit which chromosome to use, default is "chr#" to use all chromosomes'-w' or '--winsize' : what window size to use, default is 1000 '-m' or '--mincpg': force generate from window to have a minimum number of CpGs, default is 10 '-n' or '--nsample': number of samples to generate per chromosome '-p' or '--output': samples output folder, default is "out" Use script "generate_samples_concat.py" to generate samples from long-range interacting windows such as Hi-C interactions or computed correlation.Check out the example long range file in the folder "data" to form your own long-range interacting windows for sample generation and concatenation. 4.3 Training Script Use diffusion.py to train and test a DDPM model using the generated samples'-t' or '--train_folder' : the folder containing the training samples'-f' or '--model_folder' : the model folder, will be created if it does not exist'-w' or '--win_size' : window size of each sample, default is 1000'-c' or '--channel': channel size of each sample'-d' or '--cuda_device' : if you have multiple cuda gpus, select which gpu to use, default is 0"-e" or "--epoch" : how many epochs for training, default is 2000"-s" or "--earlystop" : whether to use "early stopping" during training, default is False"-p" or "--patience" : patience for early stopping, default is 10 4.4 Imputation Use diffusion_inpainting.py to perform imputation on generated samples.'-t' or '--test_folder' : the folder containing samples for imputation'-o' or '--out_folder': imputed output folder name, default="inpainting_out"'-w' or '--win_size' : window size of each sample, default is 1000'-c' or '--channel': channel size of each sample'-d' or '--cuda_device' : if you have multiple cuda gpus, select which gpu to use, default is 0 Team If you have any questions or concerns about the project, please contact the following team member: Fengyao Yan fxy134@miami.edu
Language Models are Few-Shot Learners
Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general.
Zero-Shot Text-to-Image Generation
Text-to-image generation has traditionally focused on finding better modeling assumptions for training on a fixed dataset. These assumptions might involve complex architectures, auxiliary losses, or side information such as object part labels or segmentation masks supplied during training. We describe a simple approach for this task based on a transformer that autoregressively models the text and image tokens as a single stream of data. With sufficient data and scale, our approach is competitive with previous domain-specific models when evaluated in a zero-shot fashion.
Trustworthy LLMs: a Survey and Guideline for Evaluating Large Language Models' Alignment
Ensuring alignment, which refers to making models behave in accordance with human intentions [1,2], has become a critical task before deploying large language models (LLMs) in real-world applications. For instance, OpenAI devoted six months to iteratively aligning GPT-4 before its release [3]. However, a major challenge faced by practitioners is the lack of clear guidance on evaluating whether LLM outputs align with social norms, values, and regulations. This obstacle hinders systematic iteration and deployment of LLMs. To address this issue, this paper presents a comprehensive survey of key dimensions that are crucial to consider when assessing LLM trustworthiness. The survey covers seven major categories of LLM trustworthiness: reliability, safety, fairness, resistance to misuse, explainability and reasoning, adherence to social norms, and robustness. Each major category is further divided into several sub-categories, resulting in a total of 29 sub-categories. Additionally, a subset of 8 sub-categories is selected for further investigation, where corresponding measurement studies are designed and conducted on several widely-used LLMs. The measurement results indicate that, in general, more aligned models tend to perform better in terms of overall trustworthiness. However, the effectiveness of alignment varies across the different trustworthiness categories considered. This highlights the importance of conducting more fine-grained analyses, testing, and making continuous improvements on LLM alignment. By shedding light on these key dimensions of LLM trustworthiness, this paper aims to provide valuable insights and guidance to practitioners in the field. Understanding and addressing these concerns will be crucial in achieving reliable and ethically sound deployment of LLMs in various applications.
Channel Attention Is All You Need for Video Frame Interpolation
Prevailing video frame interpolation techniques rely heavily on optical flow estimation and require additional model complexity and computational cost; it is also susceptible to error propagation in challenging scenarios with large motion and heavy occlusion. To alleviate the limitation, we propose a simple but effective deep neural network for video frame interpolation, which is end-to-end trainable and is free from a motion estimation network component. Our algorithm employs a special feature reshaping operation, referred to as PixelShuffle, with a channel attention, which replaces the optical flow computation module. The main idea behind the design is to distribute the information in a feature map into multiple channels and extract motion information by attending the channels for pixel-level frame synthesis. The model given by this principle turns out to be effective in the presence of challenging motion and occlusion. We construct a comprehensive evaluation benchmark and demonstrate that the proposed approach achieves outstanding performance compared to the existing models with a component for optical flow computation.
Deliberative Alignment: Reasoning Enables Safer Language Models
As large-scale language models increasingly impact safety-critical domains, ensuring their reliable adherence to well-defined principles remains a fundamental challenge. We introduce Deliberative Alignment, a new paradigm that directly teaches the model safety specifications and trains it to explicitly recall and accurately reason over the specifications before answering. We used this approach to align OpenAI’s o-series models [1], and achieved highly precise adherence to OpenAI’s safety policies, without requiring human-written chain-of-thoughts or answers. Deliberative Alignment pushes the Pareto frontier by simultaneously increasing robustness to jailbreaks while decreasing overrefusal rates, and also improves out-of-distribution generalization. We demonstrate that reasoning over explicitly specified policies enables more scalable, trustworthy, and interpretable alignment.
Tensor Product Attention Is All You Need
Scaling language models to handle longer input sequences typically necessitates large key-value (KV) caches, resulting in substantial memory overhead during inference. In this paper, we propose Tensor Product Attention (TPA), a novel attention mechanism that uses tensor decompositions to represent queries, keys, and values compactly, substantially shrinking the KV cache size at inference time. By factorizing these representations into contextual low-rank components and seamlessly integrating with Rotary Position Embedding (RoPE), TPA achieves improved model quality alongside memory efficiency. Based on TPA, we introduce the Tensor ProducT ATTenTion Transformer (T6), a new model architecture for sequence modeling. Through extensive empirical evaluation on language modeling tasks, we demonstrate that T6 surpasses or matches the performance of standard Transformer baselines including Multi-Head Attention (MHA), Multi-Query Attention (MQA), Grouped-Query Attention (GQA), and Multi-Head Latent Attention (MLA) across various metrics, including perplexity and a range of established evaluation benchmarks. Notably, TPA's memory efficiency and computational efficiency at decoding stage enables processing longer sequences under fixed resource constraints, addressing a critical scalability challenge in modern language models. Project Page: https://github.com/tensorgi/TPA.
Bergeron: Combating Adversarial Attacks through a Conscience-Based Alignment Framework
Research into AI alignment has grown considerably since the recent introduction of increasingly capable Large Language Models (LLMs). Unfortunately, modern methods of alignment still fail to fully prevent harmful responses when models are deliberately attacked. Such vulnerabilities can lead to LLMs being manipulated into generating hazardous content: from instructions for creating dangerous materials to inciting violence or endorsing unethical behaviors. To help mitigate this issue, we introduce Bergeron: a framework designed to improve the robustness of LLMs against attacks without any additional parameter fine-tuning. Bergeron is organized into two tiers; with a secondary LLM acting as a guardian to the primary LLM. This framework better safeguards the primary model against incoming attacks while monitoring its output for any harmful content. Empirical analysis reviews that by using Bergeron to complement models with existing alignment training, we can significantly improve the robustness and safety of multiple, commonly used commercial and open-source LLMs. Specifically, we found that models integrated with Bergeron are, on average, nearly seven times more resistant to attacks compared to models without such support.
Think in Safety: Unveiling and Mitigating Safety Alignment Collapse in Multimodal Large Reasoning Model
The rapid development of Multimodal Large Reasoning Models (MLRMs) has demonstrated broad application potential, yet their safety and reliability remain critical concerns that require systematic exploration. To address this gap, we conduct a comprehensive and systematic safety evaluation of 11 MLRMs across 5 benchmarks and unveil prevalent safety degradation phenomena in most advanced models. Moreover, our analysis reveals distinct safety patterns across different benchmarks: significant safety degradation is observed across jailbreak robustness benchmarks, whereas safety-awareness benchmarks demonstrate less pronounced degradation. In particular, the long thought process in some scenarios even enhances safety performance. Therefore, it is a potential approach to address safety issues in MLRMs by leveraging the intrinsic reasoning capabilities of the model to detect unsafe intent. To operationalize this insight, we construct a multimodal tuning dataset that incorporates a safety-oriented thought process. Experimental results from fine-tuning existing MLRMs with this dataset effectively enhances the safety on both jailbreak robustness and safety-awareness benchmarks. This study provides a new perspective for developing safe MLRMs. Our dataset is available at https://github.com/xinyuelou/Think-in-Safety.
Breach By A Thousand Leaks: Unsafe Information Leakage in 'Safe' AI Responses
Vulnerability of Frontier language models to misuse and jailbreaks has prompted the development of safety measures like filters and alignment training in an effort to ensure safety through robustness to adversarially crafted prompts. We assert that robustness is fundamentally insufficient for ensuring safety goals, and current defenses and evaluation methods fail to account for risks of dual-intent queries and their composition for malicious goals. To quantify these risks, we introduce a new safety evaluation framework based on impermissible information leakage of model outputs and demonstrate how our proposed question-decomposition attack can extract dangerous knowledge from a censored LLM more effectively than traditional jailbreaking. Underlying our proposed evaluation method is a novel information-theoretic threat model of inferential adversaries, distinguished from security adversaries, such as jailbreaks, in that success is measured by inferring impermissible knowledge from victim outputs as opposed to forcing explicitly impermissible outputs from the victim. Through our information-theoretic framework, we show that to ensure safety against inferential adversaries, defense mechanisms must ensure information censorship, bounding the leakage of impermissible information. However, we prove that such defenses inevitably incur a safety-utility trade-off.
Ensuring Safety and Trust: Analyzing the Risks of Large Language Models in Medicine
The remarkable capabilities of Large Language Models (LLMs) make them increasingly compelling for adoption in real-world healthcare applications. However, the risks associated with using LLMs in medical applications have not been systematically characterized. We propose using five key principles for safe and trustworthy medical AI – Truthfulness, Resilience, Fairness, Robustness, and Privacy – along with ten specific aspects. Under this comprehensive framework, we introduce a novel MedGuard benchmark with 1,000 expert-verified questions. Our evaluation of 11 commonly used LLMs shows that the current language models, regardless of their safety alignment mechanisms, generally perform poorly on most of our benchmarks, particularly when compared to the high performance of human physicians. Despite recent reports indicate that advanced LLMs like ChatGPT can match or even exceed human performance in various medical tasks, this study underscores a significant safety gap, highlighting the crucial need for human oversight and the implementation of AI safety guardrails.
Attention Is All You Need for KV Cache in Diffusion LLMs
This work studies how to adaptively recompute key-value (KV) caches for diffusion large language models (DLMs) to maximize prediction accuracy while minimizing decoding latency. Prior methods'decoders recompute QKV for all tokens at every denoising step and layer, despite KV states changing little across most steps, especially in shallow layers, leading to substantial redundancy. We make three observations: (1) distant ${\bf MASK}$ tokens primarily act as a length-bias and can be cached block-wise beyond the active prediction window; (2) KV dynamics increase with depth, suggesting that selective refresh starting from deeper layers is sufficient; and (3) the most-attended token exhibits the smallest KV drift, providing a conservative lower bound on cache change for other tokens. Building on these, we propose ${\bf Elastic-Cache}$, a training-free, architecture-agnostic strategy that jointly decides ${when}$ to refresh (via an attention-aware drift test on the most-attended token) and ${where}$ to refresh (via a depth-aware schedule that recomputes from a chosen layer onward while reusing shallow-layer caches and off-window MASK caches). Unlike fixed-period schemes, Elastic-Cache performs adaptive, layer-aware cache updates for diffusion LLMs, reducing redundant computation and accelerating decoding with negligible loss in generation quality. Experiments on LLaDA-Instruct, LLaDA-1.5, and LLaDA-V across mathematical reasoning and code generation tasks demonstrate consistent speedups: $8.7\times$ on GSM8K (256 tokens), and $45.1\times$ on longer sequences, while consistently maintaining higher accuracy than the baseline. Our method achieves significantly higher throughput ($6.8\times$ on GSM8K) than existing confidence-based approaches while preserving generation quality, enabling practical deployment of diffusion LLMs.
Not All Attention Is All You Need
Beyond the success story of pre-trained language models (PrLMs) in recent natural language processing, they are susceptible to over-fitting due to unusual large model size. To this end, dropout serves as a therapy. However, existing methods like random-based, knowledge-based and search-based dropout are more general but less effective onto self-attention based models, which are broadly chosen as the fundamental architecture of PrLMs. In this paper, we propose a novel dropout method named AttendOut to let self-attention empowered PrLMs capable of more robust task-specific tuning. We demonstrate that state-of-the-art models with elaborate training design may achieve much stronger results. We verify the universality of our approach on extensive natural language processing tasks.
When AIs Judge AIs: The Rise of Agent-as-a-Judge Evaluation for LLMs
As large language models (LLMs) grow in capability and autonomy, evaluating their outputs-especially in open-ended and complex tasks-has become a critical bottleneck. A new paradigm is emerging: using AI agents as the evaluators themselves. This"agent-as-a-judge"approach leverages the reasoning and perspective-taking abilities of LLMs to assess the quality and safety of other models, promising calable and nuanced alternatives to human evaluation. In this review, we define the agent-as-a-judge concept, trace its evolution from single-model judges to dynamic multi-agent debate frameworks, and critically examine their strengths and shortcomings. We compare these approaches across reliability, cost, and human alignment, and survey real-world deployments in domains such as medicine, law, finance, and education. Finally, we highlight pressing challenges-including bias, robustness, and meta evaluation-and outline future research directions. By bringing together these strands, our review demonstrates how agent-based judging can complement (but not replace) human oversight, marking a step toward trustworthy, scalable evaluation for next-generation LLMs.
Smoothed Embeddings for Robust Language Models
Improving the safety and reliability of large language models (LLMs) is a crucial aspect of realizing trustworthy AI systems. Although alignment methods aim to suppress harmful content generation, LLMs are often still vulnerable to jailbreaking attacks that employ adversarial inputs that subvert alignment and induce harmful outputs. We propose the Randomized Embedding Smoothing and Token Aggregation (RESTA) defense, which adds random noise to the embedding vectors and performs aggregation during the generation of each output token, with the aim of better preserving semantic information. Our experiments demonstrate that our approach achieves superior robustness versus utility tradeoffs compared to the baseline defenses.
Guardians of the Agentic System: Preventing Many Shots Jailbreak with Agentic System
The autonomous AI agents using large language models can create undeniable values in all span of the society but they face security threats from adversaries that warrants immediate protective solutions because trust and safety issues arise. Considering the many-shot jailbreaking and deceptive alignment as some of the main advanced attacks, that cannot be mitigated by the static guardrails used during the supervised training, points out a crucial research priority for real world robustness. The combination of static guardrails in dynamic multi-agent system fails to defend against those attacks. We intend to enhance security for LLM-based agents through the development of new evaluation frameworks which identify and counter threats for safe operational deployment. Our work uses three examination methods to detect rogue agents through a Reverse Turing Test and analyze deceptive alignment through multi-agent simulations and develops an anti-jailbreaking system by testing it with GEMINI 1.5 pro and llama-3.3-70B, deepseek r1 models using tool-mediated adversarial scenarios. The detection capabilities are strong such as 94\% accuracy for GEMINI 1.5 pro yet the system suffers persistent vulnerabilities when under long attacks as prompt length increases attack success rates (ASR) and diversity metrics become ineffective in prediction while revealing multiple complex system faults. The findings demonstrate the necessity of adopting flexible security systems based on active monitoring that can be performed by the agents themselves together with adaptable interventions by system admin as the current models can create vulnerabilities that can lead to the unreliable and vulnerable system. So, in our work, we try to address such situations and propose a comprehensive framework to counteract the security issues.
Master GAN: Multiple Attention is all you Need: A Multiple Attention Guided Super Resolution Network for Dems
The task of transforming low-resolution remote sensing images to high-resolution has consistently presented a formidable challenge in the field. The use of Generative Adversarial Networks (GANs) has led to tremendous development in the field. In this study, a novel super resolution architecture Multiple Attention Swin Transformer Enhanced Residual GAN (MASTER GAN) has been introduced, that uses multiple attention techniques in a neural network trained in an adversarial training environment. The introduced MASTER GAN acheives state-of-the-art results in super resolution tasks, when compared to existing mechanism. The paper also introduces an open source database of low resolution and counter high resolution imagery, generated using Kernel GAN. The training code has been provided at: https://github.com/sheikhazhanmohammed/MASTERGAN.git
Attention is All You Need to Defend Against Indirect Prompt Injection Attacks in LLMs
Large Language Models (LLMs) have been integrated into many applications (e.g., web agents) to perform more sophisticated tasks. However, LLM-empowered applications are vulnerable to Indirect Prompt Injection (IPI) attacks, where instructions are injected via untrustworthy external data sources. This paper presents Rennervate, a defense framework to detect and prevent IPI attacks. Rennervate leverages attention features to detect the covert injection at a fine-grained token level, enabling precise sanitization that neutralizes IPI attacks while maintaining LLM functionalities. Specifically, the token-level detector is materialized with a 2-step attentive pooling mechanism, which aggregates attention heads and response tokens for IPI detection and sanitization. Moreover, we establish a fine-grained IPI dataset, FIPI, to be open-sourced to support further research. Extensive experiments verify that Rennervate outperforms 15 commercial and academic IPI defense methods, achieving high precision on 5 LLMs and 6 datasets. We also demonstrate that Rennervate is transferable to unseen attacks and robust against adaptive adversaries.
SafeTuneBed: A Toolkit for Benchmarking LLM Safety Alignment in Fine-Tuning
As large language models (LLMs) become ubiquitous, parameter-efficient fine-tuning methods and safety-first defenses have proliferated rapidly. However, the number of approaches and their recent increase have resulted in diverse evaluations-varied datasets, metrics, and inconsistent threat settings-making it difficult to fairly compare safety, utility, and robustness across methods. To address this, we introduce SafeTuneBed, a benchmark and toolkit unifying fine-tuning and defense evaluation. SafeTuneBed (i) curates a diverse repository of multiple fine-tuning datasets spanning sentiment analysis, question-answering, multi-step reasoning, and open-ended instruction tasks, and allows for the generation of harmful-variant splits; (ii) enables integration of state-of-the-art defenses, including alignment-stage immunization, in-training safeguards, and post-tuning repair; and (iii) provides evaluators for safety (attack success rate, refusal consistency) and utility. Built on Python-first, dataclass-driven configs and plugins, SafeTuneBed requires minimal additional code to specify any fine-tuning regime, defense method, and metric suite, while ensuring end-to-end reproducibility. We showcase its value by benchmarking representative defenses across varied poisoning scenarios and tasks. By standardizing data, code, and metrics, SafeTuneBed is the first focused toolkit of its kind to accelerate rigorous and comparable research in safe LLM fine-tuning. Code is available at: https://github.com/criticalml-uw/SafeTuneBed
Graph Structure from Point Clouds: Geometric Attention is All You Need
The use of graph neural networks has produced significant advances in point cloud problems, such as those found in high energy physics. The question of how to produce a graph structure in these problems is usually treated as a matter of heuristics, employing fully connected graphs or K-nearest neighbors. In this work, we elevate this question to utmost importance as the Topology Problem. We propose an attention mechanism that allows a graph to be constructed in a learned space that handles geometrically the flow of relevance, providing one solution to the Topology Problem. We test this architecture, called GravNetNorm, on the task of top jet tagging, and show that it is competitive in tagging accuracy, and uses far fewer computational resources than all other comparable models.
Attention is all you need for an improved CNN-based flash flood susceptibility modeling. The case of the ungauged Rheraya watershed, Morocco
Effective flood hazard management requires evaluating and predicting flash flood susceptibility. Convolutional neural networks (CNNs) are commonly used for this task but face issues like gradient explosion and overfitting. This study explores the use of an attention mechanism, specifically the convolutional block attention module (CBAM), to enhance CNN models for flash flood susceptibility in the ungauged Rheraya watershed, a flood prone region. We used ResNet18, DenseNet121, and Xception as backbone architectures, integrating CBAM at different locations. Our dataset included 16 conditioning factors and 522 flash flood inventory points. Performance was evaluated using accuracy, precision, recall, F1-score, and the area under the curve (AUC) of the receiver operating characteristic (ROC). Results showed that CBAM significantly improved model performance, with DenseNet121 incorporating CBAM in each convolutional block achieving the best results (accuracy = 0.95, AUC = 0.98). Distance to river and drainage density were identified as key factors. These findings demonstrate the effectiveness of the attention mechanism in improving flash flood susceptibility modeling and offer valuable insights for disaster management.
Element-wise Attention Is All You Need
The self-attention (SA) mechanism has demonstrated superior performance across various domains, yet it suffers from substantial complexity during both training and inference. The next-generation architecture, aiming at retaining the competitive performance of SA while achieving low-cost inference and efficient long-sequence training, primarily focuses on three approaches: linear attention, linear RNNs, and state space models. Although these approaches achieve reduced complexity than SA, they all have built-in performance degradation factors, such as diminishedâspikinessâand compression of historical information. In contrast to these approaches, we propose a novel element-wise attention mechanism, which uses the element-wise squared Euclidean distance, instead of the dot product operation, to compute similarity and approximates the quadratic complexity term $\exp(q_{ic}k_{jc})$ with a Taylor polynomial. This design achieves remarkable efficiency: during training, the element-wise attention has a complexity of $\mathcal{O}(tLD)$, making long-sequence training both computationally and memory efficient, where $L$ is the sequence length, $D$ is the feature dimension, and $t$ is the highest order of the polynomial; during inference, it can be reformulated as recurrent neural networks, achieving a inference complexity of $\mathcal{O}(tD)$. Furthermore, the element-wise attention circumvents the performance degradation factors present in these approaches and achieves performance comparable to SA in both causal and non-causal forms.
Diagnosing Hallucination Risk in AI Surgical Decision-Support: A Sequential Framework for Sequential Validation
Large language models (LLMs) offer transformative potential for clinical decision support in spine surgery but pose significant risks through hallucinations, which are factually inconsistent or contextually misaligned outputs that may compromise patient safety. This study introduces a clinician-centered framework to quantify hallucination risks by evaluating diagnostic precision, recommendation quality, reasoning robustness, output coherence, and knowledge alignment. We assessed six leading LLMs across 30 expert-validated spinal cases. DeepSeek-R1 demonstrated superior overall performance (total score: 86.03 $\pm$ 2.08), particularly in high-stakes domains such as trauma and infection. A critical finding reveals that reasoning-enhanced model variants did not uniformly outperform standard counterparts: Claude-3.7-Sonnet's extended thinking mode underperformed relative to its standard version (80.79 $\pm$ 1.83 vs. 81.56 $\pm$ 1.92), indicating extended chain-of-thought reasoning alone is insufficient for clinical reliability. Multidimensional stress-testing exposed model-specific vulnerabilities, with recommendation quality degrading by 7.4% under amplified complexity. This decline contrasted with marginal improvements in rationality (+2.0%), readability (+1.7%) and diagnosis (+4.7%), highlighting a concerning divergence between perceived coherence and actionable guidance. Our findings advocate integrating interpretability mechanisms (e.g., reasoning chain visualization) into clinical workflows and establish a safety-aware validation framework for surgical LLM deployment.
Attention Is All You Need But You Don't Need All Of It For Inference of Large Language Models
The inference demand for LLMs has skyrocketed in recent months, and serving models with low latencies remains challenging due to the quadratic input length complexity of the attention layers. In this work, we investigate the effect of dropping MLP and attention layers at inference time on the performance of Llama-v2 models. We find that dropping dreeper attention layers only marginally decreases performance but leads to the best speedups alongside dropping entire layers. For example, removing 33\% of attention layers in a 13B Llama2 model results in a 1.8\% drop in average performance over the OpenLLM benchmark. We also observe that skipping layers except the latter layers reduces performances for more layers skipped, except for skipping the attention layers.
Strong Preferences Affect the Robustness of Preference Models and Value Alignment
Value alignment, which aims to ensure that large language models (LLMs) and other AI agents behave in accordance with human values, is critical for ensuring safety and trustworthiness of these systems. A key component of value alignment is the modeling of human preferences as a representation of human values. In this paper, we investigate the robustness of value alignment by examining the sensitivity of preference models. Specifically, we ask: how do changes in the probabilities of some preferences affect the predictions of these models for other preferences? To answer this question, we theoretically analyze the robustness of widely used preference models by examining their sensitivities to minor changes in preferences they model. Our findings reveal that, in the Bradley-Terry and the Placket-Luce model, the probability of a preference can change significantly as other preferences change, especially when these preferences are dominant (i.e., with probabilities near 0 or 1). We identify specific conditions where this sensitivity becomes significant for these models and discuss the practical implications for the robustness and safety of value alignment in AI systems.
LSR: Linguistic Safety Robustness Benchmark for Low-Resource West African Languages
Safety alignment in large language models relies predominantly on English-language training data. When harmful intent is expressed in low-resource languages, refusal mechanisms that hold in English frequently fail to activate. We introduce LSR (Linguistic Safety Robustness), the first systematic benchmark for measuring cross-lingual refusal degradation in West African languages: Yoruba, Hausa, Igbo, and Igala. LSR uses a dual-probe evaluation protocol - submitting matched English and target-language probes to the same model - and introduces Refusal Centroid Drift (RCD), a metric that quantifies how much of a model's English refusal behavior is lost when harmful intent is encoded in a target language. We evaluate Gemini 2.5 Flash across 14 culturally grounded attack probes in four harm categories. English refusal rates hold at approximately 90 percent. Across West African languages, refusal rates fall to 35-55 percent, with Igala showing the most severe degradation (RCD = 0.55). LSR is implemented in the Inspect AI evaluation framework and is available as a PR-ready contribution to the UK AISI's inspect_evals repository. A live reference implementation and the benchmark dataset are publicly available.
Overview of PAN 2026: Voight-Kampff Generative AI Detection, Text Watermarking, Multi-Author Writing Style Analysis, Generative Plagiarism Detection, and Reasoning Trajectory Detection
The goal of the PAN workshop is to advance computational stylometry and text forensics via objective and reproducible evaluation. In 2026, we run the following five tasks: (1) Voight-Kampff Generative AI Detection, particularly in mixed and obfuscated authorship scenarios, (2) Text Watermarking, a new task that aims to find new and benchmark the robustness of existing text watermarking schemes, (3) Multi-author Writing Style Analysis, a continued task that aims to find positions of authorship change, (4) Generative Plagiarism Detection, a continued task that targets source retrieval and text alignment between generated text and source documents, and (5) Reasoning Trajectory Detection, a new task that deals with source detection and safety detection of LLM-generated or human-written reasoning trajectories. As in previous years, PAN invites software submissions as easy-to-reproduce Docker containers for most of the tasks. Since PAN 2012, more than 1,100 submissions have been made this way via the TIRA experimentation platform.
An Analysis of Symmetry in Quantitative Semantics
In this paper, we build on a recent bicategorical model called thin spans of groupoids, introduced by Clairambault and Forest. Notably, thin spans feature a decomposition of symmetry into two sub-groupoids of polarized -- positive and negative -- symmetries. We first construct a variation of the original exponential of thin spans, based on sequences rather than families. Then we give a syntactic characterisation of the interpretation of simply-typed lambda-terms in thin spans, in terms of rigid intersection types and rigid resource terms. Finally, we formally relate thin spans with the weighted relational model and generalized species of structure. This allows us to show how some quantities in those models reflect polarized symmetries: in particular we show that the weighted relational model counts witnesses from generalized species of structure, divided by the cardinal of a group of positive symmetries.
Large Language Models and Multimodal Retrieval for Visual Word Sense Disambiguation
Visual Word Sense Disambiguation (VWSD) is a novel challenging task with the goal of retrieving an image among a set of candidates, which better represents the meaning of an ambiguous word within a given context. In this paper, we make a substantial step towards unveiling this interesting task by applying a varying set of approaches. Since VWSD is primarily a text-image retrieval task, we explore the latest transformer-based methods for multimodal retrieval. Additionally, we utilize Large Language Models (LLMs) as knowledge bases to enhance the given phrases and resolve ambiguity related to the target word. We also study VWSD as a unimodal problem by converting to text-to-text and image-to-image retrieval, as well as question-answering (QA), to fully explore the capabilities of relevant models. To tap into the implicit knowledge of LLMs, we experiment with Chain-of-Thought (CoT) prompting to guide explainable answer generation. On top of all, we train a learn to rank (LTR) model in order to combine our different modules, achieving competitive ranking results. Extensive experiments on VWSD demonstrate valuable insights to effectively drive future directions.
Mapping historical forest biomass for stock-change assessments at parcel to landscape scales
Understanding historical forest dynamics, specifically changes in forest biomass and carbon stocks, has become critical for assessing current forest climate benefits and projecting future benefits under various policy, regulatory, and stewardship scenarios. Carbon accounting frameworks based exclusively on national forest inventories are limited to broad-scale estimates, but model-based approaches that combine these inventories with remotely sensed data can yield contiguous fine-resolution maps of forest biomass and carbon stocks across landscapes over time. Here we describe a fundamental step in building a map-based stock-change framework: mapping historical forest biomass at fine temporal and spatial resolution (annual, 30m) across all of New York State (USA) from 1990 to 2019, using freely available data and open-source tools. Using Landsat imagery, US Forest Service Forest Inventory and Analysis (FIA) data, and off-the-shelf LiDAR collections we developed three modeling approaches for mapping historical forest aboveground biomass (AGB): training on FIA plot-level AGB estimates (direct), training on LiDAR-derived AGB maps (indirect), and an ensemble averaging predictions from the direct and indirect models. Model prediction surfaces (maps) were tested against FIA estimates at multiple scales. All three approaches produced viable outputs, yet tradeoffs were evident in terms of model complexity, map accuracy, saturation, and fine-scale pattern representation. The resulting map products can help identify where, when, and how forest carbon stocks are changing as a result of both anthropogenic and natural drivers alike. These products can thus serve as inputs to a wide range of applications including stock-change assessments, monitoring reporting and verification frameworks, and prioritizing parcels for protection or enrollment in improved management programs.
Assessing and mitigating systematic errors in forest attribute maps utilizing harvester and airborne laser scanning data
Cut-to-length harvesters collect useful information for modeling relationships between forest attributes and airborne laser scanning (ALS) data. However, harvesters operate in mature forests, which may introduce selection biases that can result in systematic errors in harvester data-based forest attribute maps. We fitted regression models (harvester models) for volume (V), height (HL), stem frequency (N), above-ground biomass, basal area, and quadratic mean diameter (QMD) using harvester and ALS data. Performances of the harvester models were evaluated using national forest inventory plots in an 8.7 Mha study area. We estimated biases of large-area synthetic estimators and compared efficiencies of model-assisted (MA) estimators with field data-based direct estimators. The harvester models performed better in productive than unproductive forests, but systematic errors occurred in both. The use of MA estimators resulted in efficiency gains that were largest for HL (relative efficiency, RE=6.0) and smallest for QMD (RE=1.5). The bias of the synthetic estimator was largest for N (39%) and smallest for V (1%). The latter was due to an overestimation of deciduous and an underestimation of spruce forests that by chance balanced. We conclude that a probability sample of reference observations may be required to ensure the unbiasedness of estimators utilizing harvester data.
Acute Lymphoblastic Leukemia Detection Using Hypercomplex-Valued Convolutional Neural Networks
This paper features convolutional neural networks defined on hypercomplex algebras applied to classify lymphocytes in blood smear digital microscopic images. Such classification is helpful for the diagnosis of acute lymphoblast leukemia (ALL), a type of blood cancer. We perform the classification task using eight hypercomplex-valued convolutional neural networks (HvCNNs) along with real-valued convolutional networks. Our results show that HvCNNs perform better than the real-valued model, showcasing higher accuracy with a much smaller number of parameters. Moreover, we found that HvCNNs based on Clifford algebras processing HSV-encoded images attained the highest observed accuracies. Precisely, our HvCNN yielded an average accuracy rate of 96.6% using the ALL-IDB2 dataset with a 50% train-test split, a value extremely close to the state-of-the-art models but using a much simpler architecture with significantly fewer parameters.