Search: “gpt”

LLM-as-a-judge validity in physics assessment depends more on the task than the model

arXiv:2603.14732v2 Announce Type: replace-cross Abstract: As large language models (LLMs) are increasingly considered for automated assessment and feedback, understanding when LLM marking is valid is

Content type

All Blog X Post Paper YouTube Reddit GitHub

Clear filters

Model Releases

Modality-Driven Search with Holistic Trace Judging for ARC-AGI-2

DGX agent

arXiv:2606.31543v1 Announce Type: new Abstract: Large language models can produce fluent, internally coherent reasoning traces for abstract reasoning tasks while still being confidently wrong - making

model-releasesarxiv-cs-ai

1 Jul 2026

My prediction: the excitement for Fable 5 will wear off really fast. Reposting this to help those who will be extremely disappointed after t…

My prediction: the excitement for Fable 5 will wear off really fast. Reposting this to help those who will be extremely disappointed after they play with Fable 5 and run out of tokens or can't do much

model-releasesdair-ai--x

Seeing Through Multiple Views: Parameter-Efficient Fine-Tuning via Selective Neurons for Consistent Radiology Report Generation

arXiv:2606.31099v1 Announce Type: cross Abstract: Recent years have seen substantial advances in radiology report generation (RRG), yet existing approaches predominantly adopt direct feature fusion wh

Surrogate Fidelity: When Can Open LLMs Explain Closed Ones?

model-releasesarxiv-cs-lg

arXiv:2606.32008v1 Announce Type: new Abstract: Mechanistic interpretability (MI) requires full access to model internals, yet the APIs for most widely deployed language models at best expose log-prob

This is exactly why we believe in customization. Quick context: Factory's original secret scanner was deterministic, so it either flagged th…

model-releasesfireworks-ai--x

This is exactly why we believe in customization. Quick context: Factory's original secret scanner was deterministic, so it either flagged things that weren't actually secrets (false positives) or miss

Understanding and Evaluating Claw-like Agent Security Through a Computer-Systems Lens

arXiv:2606.30755v1 Announce Type: cross Abstract: Claw-like AI agents (e.g., OpenClaw) are always-on processes with persistent access to credentials, files, tools, and external services. They take on

Local Ai

When transformers learn 'impossible' languages, what do they learn?

arXiv:2606.30815v1 Announce Type: cross Abstract: Recent work suggests that transformer language models show a bias towards human languages over unnatural ('impossible') languages argued to be unacqui

local-aiarxiv-cs-ai

Wisdom Of The (AI) Crowd: Investigating Artificial Swarm Intelligence In Large Language Models

arXiv:2606.31404v1 Announce Type: new Abstract: Human swarm intelligence demonstrates remarkable collective accuracy but faces scalability constraints in cost, coordination, and time. We investigate w

An AI agent for treatment reasoning over a biomedical tool universe

arXiv:2606.28692v1 Announce Type: new Abstract: Treatment reasoning underpins every therapeutic decision, integrating disease context, comorbidities, medications, contraindications, and evolving biome

Attractor States Emerge in Multi-Turn LLM Conversations

arXiv:2606.30571v1 Announce Type: cross Abstract: Large language models (LLMs) are increasingly used in open-ended multi-agent settings, but the long-run dynamics of model--model interaction remain po

Building to the Test: Coding Agents Deliver What You Check, Not What You Requested

arXiv:2606.28430v1 Announce Type: cross Abstract: Benchmarks are widely used to evaluate task completion by Large Language Models (LLMs), but this approach has accumulated construction-validity proble

Can Fine-Tuning Erase Your Edits? On the Fragile Coexistence of Knowledge Editing and Adaptation

arXiv:2511.05852v4 Announce Type: replace-cross Abstract: Knowledge editing (KE) offers a lightweight alternative to retraining for updating large language models (LLMs). Meanwhile, fine-tuning remain

Capability Gates Are Not Authorization: Confused-Deputy Failures in LLM Agent Frameworks

arXiv:2606.28679v1 Announce Type: cross Abstract: Tool-using LLM agents increasingly read untrusted content while holding side-effecting tools such as payments, email, CRM, and infrastructure APIs, ye

agentsarxiv-cs-ai

Cognitive World Models for Process-Level Social Influence Evaluation

arXiv:2606.29495v1 Announce Type: new Abstract: Social influence dialogue changes user behavior by altering internal cognitive states. The central evaluation question is whether the user's beliefs, de

CostBench: Evaluating Multi-Turn Cost-Optimal Planning and Adaptation in Dynamic Environments for LLM Tool-Use Agents

arXiv:2511.02734v3 Announce Type: replace Abstract: Current evaluations of Large Language Model (LLM) agents primarily emphasize task completion, often overlooking resource efficiency and adaptability

Deterministic Decisions for High-Stakes AI. A Zero-Egress Pipeline with the Deployability of RAG and the Accuracy of Machine Learning

arXiv:2606.29280v1 Announce Type: cross Abstract: We identify intervention bias as a previously unquantified failure mode of zero-shot large-language-model (LLM) educational advisory agents: without t

safetyarxiv-cs-ai

EPIC-EuroParl-UdS: Information-Theoretic Perspectives on Translation and Interpreting

arXiv:2603.09785v3 Announce Type: replace Abstract: This paper introduces an updated and combined version of the bidirectional English-German EPIC-UdS (spoken) and EuroParl-UdS (written) corpora conta

safetyarxiv-cs-cl

EVAF: A Test-Retest Protocol for Selective Parametric Consolidation

arXiv:2606.29916v1 Announce Type: cross Abstract: Long-running language agents need mechanisms for deciding which experiences should persist after the working context is gone. Retrieval systems can re

Expert Evaluation of Clinical AI Tools on Real Point-of-Care Clinical Queries

arXiv:2606.28960v1 Announce Type: new Abstract: Physicians now pose millions of clinical questions to AI tools each week, yet these tools are evaluated largely on hypothetical or exam-style questions,

Explaining Attention with Program Synthesis

arXiv:2606.19317v2 Announce Type: replace-cross Abstract: A longstanding goal of research on interpretable deep learning is to replace opaque neural computations with human-meaningful symbolic descrip

GPC: Large-Scale Generative Pretraining for Transferable Motor Control

arXiv:2606.29148v1 Announce Type: cross Abstract: Developing controllers capable of completing a wide range of tasks in a natural and life-like manner is a key challenge in enabling practical applicat

safetyarxiv-cs-ai

How LLMs See Creativity: Zero-Shot Scoring of Visual Creativity with Interpretable Reasoning

arXiv:2606.29672v1 Announce Type: new Abstract: Evaluating the originality of visual images poses enduring challenges for creativity assessment. Automated scoring using AI models has proven effective

Labeling Training Data for Entity Matching Using Large Language Models

arXiv:2606.28823v1 Announce Type: new Abstract: Recent large language models (LLMs) achieve strong performance on entity matching without requiring task-specific training data. However, applying these

Mechanistically Eliciting Latent Behaviors in Language Models

arXiv:2606.29604v1 Announce Type: cross Abstract: We aim to discover diverse, generalizable perturbations of LLM internals that can surface hidden behavioral modes. Such perturbations could help resha

safetyarxiv-cs-ai

MemDelta: Controlled Baselines and Hidden Confounds in Agent Memory Evaluation

arXiv:2606.29914v1 Announce Type: new Abstract: Agent memory systems are increasingly evaluated against RAG and full-context baselines, but reported gains often mix changes in the memory method with c

Research

On the Nonlinearity of Learning Rate Scaling for LLM Training

arXiv:2606.29158v1 Announce Type: cross Abstract: Learning-rate transfer can reduce the cost of training large language models: instead of sweeping learning rates at target scale, practitioners extrap

researcharxiv-cs-ai

OSWorld2.0: Benchmarking Computer Use Agents on Long-Horizon Real-World Tasks

arXiv:2606.29537v1 Announce Type: new Abstract: Existing computer-use benchmarks fail to capture the realism, complexity, and long-horizon demands of real-world computer use, limiting their ability to

PolicyGuard: A Dialogue-Grounded Sub-Agent Verifier for Policy Adherence in LLM Agents

arXiv:2606.29225v1 Announce Type: new Abstract: LLM agents handle user requests on behalf of organizations through tool calls and must follow the company policies stated in their system prompts. Prior

S-Agent: Spatial Tool-Use Elicits Reasoning for Spatial Intelligence

model-releasesarxiv-cs-cv

arXiv:2606.20515v2 Announce Type: replace Abstract: Real-world spatial intelligence requires reasoning over a continuous and evolving 3D world, yet existing VLMs and tool-augmented agents largely rema

SafePyramid: A Hierarchical Benchmark for In-context Policy Guardrailing

arXiv:2606.29887v1 Announce Type: new Abstract: In real-world applications, guardrails are often expected to identify unsafe user-model interactions according to application-specific safety policies,

Selective Memory Retention for Long-Horizon LLM Agents

arXiv:2606.29178v1 Announce Type: new Abstract: When does retention matter for memory-augmented LLM agents? We study this with TraceRetain, a lightweight framework for bounded external memory in froze

SEVA: Self-Evolving Verification Agent with Process Reward for Fact Attribution

arXiv:2606.29713v1 Announce Type: cross Abstract: Hallucination is the reliability bottleneck for LLM-based agents, and fact attribution verifiers are the last line of defense -- yet today's verifiers

SFBench: The SciFy Scientific Feasibility Benchmark

arXiv:2606.29630v1 Announce Type: new Abstract: We present SFBench, a benchmark dataset for evaluating systems that assess the feasibility of scientific claims. SFBench includes 197 claims in material

StarDojo: Benchmarking Open-Ended Behaviors of Agentic Multimodal LLMs in Production-Living Simulations with Stardew Valley

arXiv:2507.07445v3 Announce Type: replace Abstract: Autonomous agents navigating human society must master both production activities and social interactions, yet existing benchmarks rarely evaluate t

Statistically Indistinguishable, Operationally Distinct: A Formal Barrier for Tabular Foundation Models

arXiv:2606.29091v1 Announce Type: cross Abstract: Tabular foundation models cannot reason about data produced by running systems without access to the rules that govern them. We make this statement fa

SWE-INTERACT: Reimagining SWE Benchmarks as User-Driven Long-Horizon Coding Sessions

arXiv:2606.30573v1 Announce Type: new Abstract: We introduce SWE-Interact, a new testbed for evaluating coding agents on multi-turn, interactive, user-driven software engineering tasks. Existing front

agentsarxiv-cs-lg

The Digital Afterlife of Empires: Four Language Models Converge on the Same Imperial Cartography of Writing

arXiv:2606.28325v1 Announce Type: cross Abstract: Large language models process the world's writing systems with radical inequality. We constructed the Digital Script Representation Index (DSRI), a se

Timesteps of Mamba Align with Human Reading Times

arXiv:2606.29904v1 Announce Type: new Abstract: This study demonstrates an alignment of per-word processing time in a popular state-space language model Mamba and human readers. In Mamba, the recurren

safetyarxiv-cs-cl

Tool-Augmented Spatiotemporal Reasoning for Streamlining Video Question Answering Task

arXiv:2512.10359v1 Announce Type: cross Abstract: Video Question Answering (VideoQA) task serves as a critical playground for evaluating whether foundation models can effectively perceive, understand,

agentsarxiv-cs-ai

An Empirical Analysis of Factual Errors in Human-Written Text and its Application

arXiv:2606.27959v1 Announce Type: new Abstract: Factual Error Detection (FED), which is the task of identifying factually incorrect spans in a given text, has long been recognized as an important rese

CalBrief: A Pilot Diagnostic Benchmark for Evidence-Calibrated Scientific Briefing with Large Language Models

arXiv:2606.27383v1 Announce Type: cross Abstract: Large language models (LLMs) are increasingly used as research assistants, yet it remains unclear whether they can calibrate research takeaways to the

Grounded Iterative Language Planning: How Parameterized World Models Reduce Hallucination Propagation in LLM Agents

arXiv:2606.27806v1 Announce Type: new Abstract: World models for language agents come in two useful forms. An agent-based world model calls an LLM API and reasons flexibly in language, but its errors

agentsarxiv-cs-ai

Multimodal Evaluator Preference Collapse: Cross-Modal Coupling in Self-Evolving Agents

arXiv:2606.16682v3 Announce Type: replace-cross Abstract: When AI agents use language models to evaluate their own outputs in a feedback loop, systematic biases emerge. We show that Evaluator Preferen

NormAct: A Benchmark for Hidden Social Norm Compliance in Embodied Planning

arXiv:2606.27826v1 Announce Type: new Abstract: Multimodal large language models (MLLMs) are increasingly deployed as embodied planners in egocentric environments, where task success requires not only

Supersede: Diagnosing and Training the Memory-Update Gap in LLM Agents

arXiv:2606.27472v1 Announce Type: cross Abstract: Large language model (LLM) agents operate over long, multi-session interactions in which facts change: a user moves, a price updates, a plan is revise

Research

The Curse of Multiple Mediators: Hidden Interaction Effects in Activation Patching

arXiv:2606.27510v1 Announce Type: cross Abstract: Activation patching is the primary tool in mechanistic interpretability. It attributes causal responsibility for a model behavior to each of its indiv

researcharxiv-cs-cl

VASAE: Naming SAE Dictionary Directions with Vocabulary-Aligned Anchoring

arXiv:2606.27941v1 Announce Type: cross Abstract: Sparse autoencoders (SAEs) provide useful decompositions of Transformer residual streams, but their learned features are usually named post hoc rather