All entries

Polysemantic Experts, Monosemantic Paths: Routing as Control in MoEs

arXiv:2604.17837v1 Announce Type: cross Abstract: An LLM's residual stream is both state and instruction: it encodes the current context and determines the next transformation. We introduce a paramete

Position: Multimodal Large Language Models Can Significantly Advance Scientific Reasoning

DGX agent

arXiv:2502.02871v2 Announce Type: replace Abstract: Scientific reasoning, the process through which humans apply logic, evidence, and critical thinking to explore and interpret scientific phenomena, i

researcharxiv-cs-cl

21 Apr 2026

Agents

Positioning radiata pine branches requiring pruning by drone stereo vision

arXiv:2604.16480v1 Announce Type: new Abstract: This paper presents a stereo-vision-based system mounted on a drone for detecting and localising radiata pine branches to support autonomous pruning. Th

agentsarxiv-cs-cv

Positive-Only Drifting Policy Optimization

arXiv:2604.16519v1 Announce Type: new Abstract: In the field of online reinforcement learning (RL), traditional Gaussian policies and flow-based methods are often constrained by their unimodal express

local-aiarxiv-cs-lg

PowerCLIP: Powerset Alignment for Contrastive Pre-Training

arXiv:2511.23170v5 Announce Type: replace Abstract: Contrastive vision-language pre-training frameworks such as CLIP have demonstrated impressive zero-shot performance across a range of vision-languag

safetyarxiv-cs-cv

PPEDCRF: Dynamic-CRF-Guided Selective Perturbation for Background-Based Location Privacy in Video Sequences

arXiv:2604.17163v1 Announce Type: new Abstract: We propose PPEDCRF, a calibrated selective perturbation framework that protects background-based location privacy in released video frames against galle

Precise Debugging Benchmark: Is Your Model Debugging or Regenerating?

arXiv:2604.17338v1 Announce Type: cross Abstract: Unlike code completion, debugging requires localizing faults and applying targeted edits. We observe that frontier LLMs often regenerate correct but o

Predicting Blastocyst Formation in IVF: Integrating DINOv2 and Attention-Based LSTM on Time-Lapse Embryo Images

arXiv:2604.16505v1 Announce Type: new Abstract: The selection of the optimal embryo for transfer is a critical yet challenging step in in vitro fertilization (IVF), primarily due to its reliance on th

researcharxiv-cs-cv

Predicting LLM Compression Degradation from Spectral Statistics

arXiv:2604.18085v1 Announce Type: new Abstract: Matrix-level low-rank compression is a promising way to reduce the cost of large language models, but running compression and evaluating the resulting m

Predictive Modeling of Natural Medicinal Compounds for Alzheimer Disease Using Cheminformatics

arXiv:2604.18316v1 Announce Type: cross Abstract: The most common cause of dementia is Alzheimer disease, a progressive neurodegenerative disorder affecting older adults that gradually impairs memory,

researcharxiv-cs-lg

Pref-GRPO: Pairwise Preference Reward-based GRPO for Stable Text-to-Image Reinforcement Learning

arXiv:2508.20751v2 Announce Type: replace Abstract: Recent advancements highlight the importance of GRPO-based reinforcement learning methods and benchmarking in enhancing text-to-image (T2I) generati

PrefixMemory-Tuning: Modernizing Prefix-Tuning by Decoupling the Prefix from Attention

arXiv:2506.13674v3 Announce Type: replace Abstract: Parameter-Efficient Fine-Tuning (PEFT) methods have become crucial for rapidly adapting large language models (LLMs) to downstream tasks. Prefix-Tun

Preparation of Fractal-Inspired Computational Architectures for Automated Neural Design Exploration

arXiv:2511.07329v3 Announce Type: replace-cross Abstract: It introduces FractalNet, a fractal-inspired computational architectures for advanced large language model analysis that mainly challenges mod

researcharxiv-cs-cv

Applications

Preventing overfitting in deep learning using differential privacy

arXiv:2604.16334v1 Announce Type: new Abstract: The use of Deep Neural Network based systems in the real world is growing. They have achieved state-of-the-art performance on many image, speech and tex

applicationsarxiv-cs-lg

PrinciplismQA: A Philosophy-Grounded Approach to Assessing LLM-Human Clinical Medical Ethics Alignment

arXiv:2508.05132v2 Announce Type: replace Abstract: As medical LLMs transition to clinical deployment, assessing their ethical reasoning capability becomes critical. While achieving high accuracy on k

safetyarxiv-cs-cl

Prior-Fitted Functional Flow: In-Context Generative Models for Pharmacokinetics

arXiv:2604.17670v1 Announce Type: new Abstract: We introduce Prior-Fitted Functional Flows, a generative foundation model for pharmacokinetics that enables zero-shot population synthesis and individua

PRISM: Probing Reasoning, Instruction, and Source Memory in LLM Hallucinations

arXiv:2604.16909v1 Announce Type: new Abstract: As large language models (LLMs) evolve from conversational assistants into agents capable of handling complex tasks, they are increasingly deployed in h

PRISMA: Preference-Reinforced Self-Training Approach for Interpretable Emotionally Intelligent Negotiation Dialogues

arXiv:2604.18354v1 Announce Type: new Abstract: Emotion plays a pivotal role in shaping negotiation outcomes, influencing trust, cooperation, and long-term relationships. Developing negotiation dialog

researcharxiv-cs-cl

Privacy Collapse: Benign Fine-Tuning Can Break Contextual Privacy in Language Models

arXiv:2601.15220v2 Announce Type: replace Abstract: We identify a novel phenomenon in language models: benign fine-tuning of frontier models can lead to privacy collapse. We find that diverse, subtle

safetyarxiv-cs-cl

Privacy-Preserving Semantic Segmentation without Key Management

arXiv:2604.16523v1 Announce Type: new Abstract: This paper proposes a novel privacy-preserving semantic segmentation method that can use independent keys for each client and image. In the proposed met

local-aiarxiv-cs-cv

Privacy-R1: Privacy-Aware Multi-LLM Agent Collaboration via Reinforcement Learning

arXiv:2510.16054v2 Announce Type: replace-cross Abstract: When users submit queries to Large Language Models (LLMs), their prompts can often contain sensitive data, forcing a difficult choice: Send th

local-aiarxiv-cs-cl

PrivaDE: Privacy-preserving Data Evaluation for Blockchain-based Data Marketplaces

arXiv:2510.18109v4 Announce Type: replace-cross Abstract: Evaluating the usefulness of data before purchase is essential when obtaining data for high-quality machine learning models, yet both model bu

researcharxiv-cs-lg

Privatar: Scalable Privacy-preserving Multi-user VR via Secure Offloading

arXiv:2604.17476v1 Announce Type: cross Abstract: Multi-user virtual reality enables immersive interaction. However, rendering avatars for numerous participants on each headset incurs prohibitive comp

local-aiarxiv-cs-cv

PRL: Prompts from Reinforcement Learning

arXiv:2505.14412v2 Announce Type: replace-cross Abstract: Effective prompt engineering remains a central challenge in fully harnessing the capabilities of LLMs. While well-designed prompts can dramati

researcharxiv-cs-cl

Hardware

Probabilistic Programs of Thought

arXiv:2604.17290v1 Announce Type: new Abstract: LLMs are widely used for code generation and mathematical reasoning tasks where they are required to generate structured output. They either need to rea

hardwarearxiv-cs-cl

Agents

probably true (and agree on it being open)

This post by Harrison Chase (creator of LangChain) likely discusses the concept of statements or propositions that are 'probably true' and advocates for keeping such discussions or determinations open

agentsharrison-chase--x

Tutorials

Procedural Knowledge at Scale Improves Reasoning

arXiv:2604.01348v2 Announce Type: replace Abstract: Test-time scaling has emerged as an effective way to improve language models on challenging reasoning tasks. However, most existing methods treat ea

tutorialsarxiv-cs-cl

Process Reward Models Meet Planning: Generating Precise and Scalable Datasets for Step-Level Rewards

arXiv:2604.17957v1 Announce Type: new Abstract: Process Reward Models (PRMs) have emerged as a powerful tool for providing step-level feedback when evaluating the reasoning of Large Language Models (L

researcharxiv-cs-cl

ProfVLM: A lightweight video-language model for multi-view proficiency estimation

arXiv:2509.26278v4 Announce Type: replace-cross Abstract: Most existing approaches formulate action quality assessment and skill proficiency estimation as discriminative prediction tasks, typically pr

Program Structure-aware Language Models: Targeted Software Testing beyond Textual Semantics

arXiv:2604.17715v1 Announce Type: cross Abstract: Recent advances in large language models for test case generation have improved branch coverage via prompt-engineered mutations. However, they still l

Progressive Online Video Understanding with Evidence-Aligned Timing and Transparent Decisions

arXiv:2604.18459v1 Announce Type: new Abstract: Visual agents operating in the wild must respond to queries precisely when sufficient evidence first appears in a video stream, a critical capability th

Projected Coupled Diffusion for Test-Time Constrained Joint Generation

arXiv:2508.10531v3 Announce Type: replace Abstract: Modifications to test-time sampling have emerged as an important extension to diffusion algorithms, with the goal of biasing the generative process

researcharxiv-cs-lg

Prompt Sensitivity in Vision-Language Grounding: How Small Changes in Wording Affect Object Detection

arXiv:2604.17126v1 Announce Type: new Abstract: Vision-language models enable open-vocabulary object grounding through natural language queries, under the implicit assumption that semantically equival

researcharxiv-cs-cv

Prompting Foundation Models for Zero-Shot Ship Instance Segmentation in SAR Imagery

arXiv:2604.17920v1 Announce Type: new Abstract: Synthetic Aperture Radar (SAR) plays a critical role in maritime surveillance, yet deep learning for SAR analysis is limited by the lack of pixel-level

Applications

Proprietary lock-in is holding back enterprise AI ambition, says SUSE CEO

As enterprises race to adopt AI without sacrificing control or flexibility, open-source infrastructure is the stable foundation that enables organizations to modernize while maintaining full digital s

applicationssiliconangle

Industry

Proprietary stacks are failing enterprises when they need flexibility the most

As hybrid and multicloud strategies evolve, a new reality is coming into focus — one where open architectures are increasingly central to operational resilience, sovereignty and choice. The industry’s

industrysiliconangle

ProtoCLIP: Prototype-Aligned Latent Refinement for Robust Zero-Shot Chest X-Ray Classification

arXiv:2604.18444v1 Announce Type: cross Abstract: Zero-shot vision-language models (VLMs) have shown promise for chest radiograph classification, but their performance is often limited by confounding

safetyarxiv-cs-cv

ProTrain: Efficient LLM Training via Memory-Aware Techniques

arXiv:2406.08334v2 Announce Type: replace-cross Abstract: Memory pressure has emerged as a dominant constraint in scaling the training of large language models (LLMs), particularly in resource-constra

Prune, Interpret, Evaluate: A Cross-Layer Transcoder-Native Framework for Efficient Circuit Discovery via Feature Attribution

arXiv:2604.16889v1 Announce Type: new Abstract: Existing feature-interpretation pipelines typically operate on uniformly sampled units, but only a small fraction of cross-layer transcoder (CLT) featur

Pseudo2Real: Task Arithmetic for Pseudo-Label Correction in Automatic Speech Recognition

arXiv:2510.08047v2 Announce Type: replace-cross Abstract: Robust ASR under domain shift is crucial because real-world systems encounter unseen accents and domains with limited labeled data. Although p

PTLD: Sim-to-real Privileged Tactile Latent Distillation for Dexterous Manipulation

model-releasesarxiv-cs-ro

arXiv:2603.04531v2 Announce Type: replace Abstract: Tactile dexterous manipulation is essential to automating complex household tasks, yet learning effective control policies remains a challenge. Whil

Tools

Publish an app in one click 💻

Replit announced a feature enabling users to publish applications with a single click, streamlining the deployment process from development to production. This capability reduces friction in the app p

toolsreplit--x

Pulse Shape Discrimination Algorithms: Survey and Benchmark

arXiv:2508.02750v2 Announce Type: replace Abstract: This review presents a comprehensive survey and benchmark of pulse shape discrimination (PSD) algorithms for radiation detection, classifying nearly

Applications

Putting a Face to Forgetting: Continual Learning meets Mechanistic Interpretability

arXiv:2601.22012v2 Announce Type: replace Abstract: Catastrophic forgetting in continual learning is often measured at the performance or last-layer representation level, overlooking the underlying me

applicationsarxiv-cs-lg

Tutorials

PyEPO: A PyTorch-based End-to-End Predict-then-Optimize Library for Linear and Integer Programming

arXiv:2206.14234v3 Announce Type: replace-cross Abstract: In deterministic optimization, it is typically assumed that all problem parameters are fixed and known. In practice, however, some parameters

tutorialsarxiv-cs-lg

Q-DeepSight: Incentivizing Thinking with Images for Image Quality Assessment and Refinement

arXiv:2604.16858v1 Announce Type: new Abstract: Image Quality Assessment (IQA) models are increasingly deployed as perceptual critics to guide generative models and image restoration. This role demand

local-aiarxiv-cs-cv

Q-SINDy: Quantum-Kernel Sparse Identification of Nonlinear Dynamics with Provable Coefficient Debiasing

arXiv:2604.16779v1 Announce Type: cross Abstract: Quantum feature maps offer expressive embeddings for classical learning tasks, and augmenting sparse identification of nonlinear dynamics (SINDy) with

safetyarxiv-cs-lg

Tools

QIMMA قِمّة ⛰: A Quality-First Arabic LLM Leaderboard

QIMMA (قِمّة) is a quality-focused leaderboard for evaluating Arabic Large Language Models, developed by TIIUAE. The leaderboard benchmarks Arabic LLMs across various performance metrics and tasks to

toolshugging-face