Search: “gpt”

arXiv:2605.14087v1 Announce Type: new Abstract: Large Language Models (LLMs), when trained on web-scale corpora, inherently absorb toxic patterns from their training data. This leads to ``toxic degene

safetyarxiv-cs-cl

Content type

All Blog X Post Paper YouTube Reddit GitHub

Clear filters

MemReranker: Reasoning-Aware Reranking for Agent Memory Retrieval

DGX agent

arXiv:2605.06132v2 Announce Type: replace Abstract: In agent memory systems, the reranking model serves as the critical bridge connecting user queries with long-term memory. Most systems adopt the 're

model-releasesarxiv-cs-cl

15 May 2026

MultiEmo-Bench: Multi-label Visual Emotion Analysis for Multi-modal Large Language Models

arXiv:2605.14635v1 Announce Type: cross Abstract: This paper introduces a multi-label visual emotion analysis benchmark dataset for comprehensively evaluating the ability of multimodal large language

MUON+: Towards More Effective Muon via One Additional Normalization Step for LLM Pre-training

arXiv:2602.21545v3 Announce Type: replace Abstract: Muon has recently emerged as a strong optimizer for large language model pre-training, orthogonalizing the momentum matrix via Newton--Schulz polar

On the Cultural Anachronism and Temporal Reasoning in Vision Language Models

arXiv:2605.15071v1 Announce Type: cross Abstract: Vision-Language Models (VLMs) are increasingly applied to cultural heritage materials, from digital archives to educational platforms. This work ident

SCOOTER: A Human Evaluation Framework for Unrestricted Adversarial Examples

arXiv:2507.07776v3 Announce Type: replace Abstract: Unrestricted adversarial attacks aim to fool computer vision models without being constrained by ell_p-norm bounds to remain imperceptible to humans

SPIN: Structural LLM Planning via Iterative Navigation for Industrial Tasks

arXiv:2605.14051v1 Announce Type: new Abstract: Industrial LLM agent systems often separate planning from execution, yet LLM planners frequently produce structurally invalid or unnecessarily long work

SVAG-Bench: A Large-Scale Benchmark for Multi-Instance Spatio-temporal Video Action Grounding

arXiv:2510.13016v3 Announce Type: replace Abstract: A truly capable AI system must do more than detect objects or recognize activities in isolation. It must form unified, grounded representations of w

Sycophancy is an Educational Safety Risk: Why LLM Tutors Need Sycophancy Benchmarks

arXiv:2605.14604v1 Announce Type: new Abstract: This position paper argues that effective tutoring requires corrective friction: surfacing misconceptions and challenging them supportively to drive con

TFGN: Task-Free, Replay-Free Continual Pre-Training Without Catastrophic Forgetting at LLM Scale

arXiv:2605.15053v1 Announce Type: cross Abstract: Continually pre-training a large language model on heterogeneous text domains, without replay or task labels, has remained an unsolved architectural p

Local Ai

Venus-DeFakerOne: Unified Fake Image Detection & Localization

arXiv:2605.14091v1 Announce Type: new Abstract: In recent years, the rapid evolution of generative AI has fundamentally reshaped the paradigm of image forgery, breaking the traditional boundaries betw

local-aiarxiv-cs-cv

VerbalValue: A Socially Intelligent Virtual Host for Sales-Driven Live Commerce

arXiv:2605.14542v1 Announce Type: new Abstract: A skilled live-commerce host is not merely a narrator, but a sales agent who converts viewer curiosity into purchase intent through expert product knowl

When Retrieval Hurts Code Completion: A Diagnostic Study of Stale Repository Context

arXiv:2605.14478v1 Announce Type: cross Abstract: Context: Retrieval-augmented code generation relies on cross-file repository context, but retrieved snippets may come from obsolete project states. Ob

When Robots Do the Chores: A Benchmark and Agent for Long-Horizon Household Task Execution

arXiv:2605.14504v1 Announce Type: new Abstract: Long-horizon household tasks demand robust high-level planning and sustained reasoning capabilities, which are largely overlooked by existing embodied A

BEAVER: An Enterprise Benchmark for Text-to-SQL

arXiv:2409.02038v3 Announce Type: replace-cross Abstract: Existing text-to-SQL benchmarks have largely been constructed from public databases with well-structured schemas and simplistic question-SQL p

Research

Continual Learning with Multilingual Foundation Model

arXiv:2605.13415v1 Announce Type: cross Abstract: This paper presents a multi-stage framework for detecting reclaimed slurs in multilingual social media discourse. It addresses the challenge of identi

researcharxiv-cs-ai

From Pixels to BFS: High Maze Accuracy Does Not Imply Visual Planning

arXiv:2603.26839v2 Announce Type: replace-cross Abstract: How do multimodal models solve visual spatial tasks -- through genuine planning, or through brute-force search in token space? We introduce ex

HLS-Seek: QoR-Aware Code Generation for High-Level Synthesis via Proxy Comparative Reward Reinforcement Learning

arXiv:2605.13536v1 Announce Type: cross Abstract: High-Level Synthesis (HLS) compiles algorithmic C/C++ descriptions into hardware, with Quality of Results (QoR) -- latency and resource utilization --

(How) Do Large Language Models Understand High-Level Message Sequence Charts?

arXiv:2605.13773v1 Announce Type: cross Abstract: Large Language Models (LLMs) are being employed widely to automate tasks across the software development life-cycle. It is, however, unclear whether t

Tutorials

Inline Critic Steers Image Editing

arXiv:2605.12724v1 Announce Type: cross Abstract: Instruction-based image editing exhibits heterogeneous difficulty not only across cases but also across regions of an image, motivating refinement app

tutorialsarxiv-cs-ai

Language Model Goal Selection Differs from Humans' in a Self-Directed Learning Task

arXiv:2603.03295v2 Announce Type: replace-cross Abstract: Whether in agentic workflows, social studies, or chat settings, large language models (LLMs) are increasingly being asked to replace humans in

LENS: Multi-level Evaluation of Multimodal Reasoning with Large Language Models

arXiv:2505.15616v2 Announce Type: replace Abstract: Multimodal Large Language Models (MLLMs) have achieved significant advances in integrating visual and linguistic information, yet their ability to r

Negation Neglect: When models fail to learn negations in training

arXiv:2605.13829v1 Announce Type: cross Abstract: We introduce Negation Neglect, where finetuning LLMs on documents that flag a claim as false makes them believe the claim is true. For example, models

PersonalAlign: Hierarchical Implicit Intent Alignment for Personalized GUI Agent with Long-Term User-Centric Records

arXiv:2601.09636v2 Announce Type: replace Abstract: While GUI agents have shown strong performance under explicit and completion instructions, real-world deployment requires aligning with users' more

Procedural Refinement by LLM-driven Algorithmic Debugging for ARC-AGI-2

arXiv:2603.20334v2 Announce Type: replace-cross Abstract: In high-complexity abstract reasoning, a system must infer a latent rule from a few examples or structured observations and apply it to unseen

RTLC -- Research, Teach-to-Learn, Critique: A three-stage prompting paradigm inspired by the Feynman Learning Technique that lifts LLM-as-judge accuracy on JudgeBench with no fine-tuning

arXiv:2605.13695v1 Announce Type: cross Abstract: LLM-as-a-judge is now the default measurement instrument for open-ended generation, but on the public JudgeBench benchmark even strong instruction-tun

SpatialReward: Bridging the Perception Gap in Online RL for Image Editing via Explicit Spatial Reasoning

arXiv:2602.07458v4 Announce Type: replace Abstract: Online Reinforcement Learning (RL) offers a promising avenue for complex image editing but is currently constrained by the scarcity of reliable and

Training Large Language Models to Predict Clinical Events

arXiv:2605.12817v1 Announce Type: cross Abstract: Longitudinal clinical notes contain rich evidence of how patients evolve over time, but converting this signal into training supervision for clinical

Unlocking Patch-Level Features for CLIP-Based Class-Incremental Learning

arXiv:2605.13835v1 Announce Type: new Abstract: Class-Incremental Learning (CIL) enables models to continuously integrate new knowledge while mitigating catastrophic forgetting. Driven by the remarkab

Useful Memories Become Faulty When Continuously Updated by LLMs

arXiv:2605.12978v1 Announce Type: new Abstract: Learning from past experience benefits from two complementary forms of memory: episodic traces -- raw trajectories of what happened -- and consolidated

VERA-MH Concept Paper

arXiv:2510.15297v4 Announce Type: replace-cross Abstract: We introduce VERA-MH (Validation of Ethical and Responsible AI in Mental Health), an automated evaluation of the safety of AI chatbots used in

Asymmetry of capabilities is when there is great risk. Diffusion of capabilities maintains an adversarial equilibrium and stability. Don't l…

Asymmetry of capabilities is when there is great risk. Diffusion of capabilities maintains an adversarial equilibrium and stability. Don't let the fear-mongerers win. We asked the CEO of HuggingFace @

Beyond Text Prompts: Visual-to-Visual Generation as A Unified Paradigm

arXiv:2605.12271v1 Announce Type: new Abstract: Humans often specify and create through visual artifacts: typography sheets, sketches, reference images, and annotated scenes. Yet modern visual generat

ClinicalBench: Stress-Testing Assertion-Aware Retrieval for Cross-Admission Clinical QA on MIMIC-IV

model-releasesarxiv-cs-cl

arXiv:2605.11143v1 Announce Type: new Abstract: Reasoning benchmarks measure clinical performance on clean inputs. We evaluate the step before reasoning: retrieval over real EHR notes, where negation,

Co-sign.

Co-sign. We asked the CEO of HuggingFace @ClementDelangue what the risks of releasing powerful open source models are. He says restricting AI creates more risk than openness. 'Six, seven years ago, at

CSP Allow-list Experiment

model-releasessimon-willison

Tool: CSP Allow-list Experiment An experiment that shows that you can load an app in a CSP-protected sandboxed iframe (see previous note) and have a custom fetch() that intercepts CSP errors and passe

ExploitGym: Can AI Agents Turn Security Vulnerabilities into Real Attacks?

arXiv:2605.11086v1 Announce Type: cross Abstract: AI agents are rapidly gaining capabilities that could significantly reshape cybersecurity, making rigorous evaluation urgent. A critical capability is

Good catch @theojaffee 😂

Good catch @theojaffee 😂 Media We asked the CEO of HuggingFace @ClementDelangue what the risks of releasing powerful open source models are. He says restricting AI creates more risk than openness. 'Si

Measuring Five-Nines Reliability: Sample-Efficient LLM Evaluation in Saturated Benchmarks

arXiv:2605.11209v1 Announce Type: new Abstract: While existing benchmarks demonstrate the near-perfect performance of large language models (LLMs) on various tasks, this apparent saturation often obsc

Local Ai

Muon is Not That Special: Random or Inverted Spectra Work Just as Well

arXiv:2605.11181v1 Announce Type: new Abstract: The recent empirical success of the Muon optimizer has renewed interest in non-Euclidean optimization, typically justified by similarities with second-o

local-aiarxiv-cs-lg

MuonQ: Enhancing Low-Bit Muon Quantization via Directional Fidelity Optimization

arXiv:2605.11396v1 Announce Type: new Abstract: The Muon optimizer has emerged as a compelling alternative to Adam for training large language models, achieving remarkable computational savings throug

Safety

Persona-Conditioned Adversarial Prompting: Multi-Identity Red-Teaming for Adversarial Discovery and Mitigation

arXiv:2605.11730v1 Announce Type: new Abstract: Automated red-teaming for LLMs often discovers narrow attack slices, missing diverse real-world threats, and yielding insufficient data for safety fine-

safetyarxiv-cs-lg

Research

Predicting Disagreement with Human Raters in LLM-as-a-Judge Difficulty Assessment without Using Generation-Time Probability Signals

arXiv:2605.12422v1 Announce Type: new Abstract: Automatic generation of educational materials using large language models (LLMs) is becoming increasingly common, but assigning difficulty levels to suc

researcharxiv-cs-cl

Procedural-skill SFT across capacity tiers: A W-Shaped pre-SFT Trajectory and Regime-Asymmetric Mechanism on 0.8B-4B Qwen3.5 Models

arXiv:2605.11907v1 Announce Type: new Abstract: We measure procedural-skill SFT contribution across three Qwen3.5 dense scales (0.8B, 2B, 4B) on a 200-task / 40-skill holdout, with Claude Haiku 4.5 as

ShapeCodeBench: A Renewable Benchmark for Perception-to-Program Reconstruction of Synthetic Shape Scenes

arXiv:2605.11680v1 Announce Type: new Abstract: We introduce ShapeCodeBench, a synthetic benchmark for perception-to-program reconstruction: given a rendered raster image, a model must emit an executa

Solve the Loop: Attractor Models for Language and Reasoning

model-releasesarxiv-cs-cl

arXiv:2605.12466v1 Announce Type: cross Abstract: Looped Transformers offer a promising alternative to purely feed-forward computation by iteratively refining latent representations, improving languag

Three Regimes of Context-Parametric Conflict: A Predictive Framework and Empirical Validation

model-releasesarxiv-cs-cl

arXiv:2605.11574v1 Announce Type: new Abstract: The literature on how large language models handle conflict between their training knowledge and a contradicting document presents a persistent empirica

We asked the CEO of HuggingFace @ClementDelangue what the risks of releasing powerful open source models are. He says restricting AI creates…

We asked the CEO of HuggingFace @ClementDelangue what the risks of releasing powerful open source models are. He says restricting AI creates more risk than openness. 'Six, seven years ago, at the time