Search: “llm”

When a Kindergartener Solves Calculus: Measuring Capability Leakage in Role-Prompted Reasoning Models

arXiv:2609.39846v1 Announce Type: cross Abstract: We investigate the problem of role-capability leakage (RCL), in which a role-prompted reasoning model generates convincing in-role text while continui

Content type

All Blog X Post Paper YouTube Reddit GitHub

Clear filters

Safety

When Clipping Reverses Correction: Failure Dynamics of Pointwise Forward-KL On-Policy Self-Distillation

DGX agent

arXiv:2609.38995v1 Announce Type: new Abstract: On-policy self-distillation (OPSD) trains a student on its own generated responses using feedback from the same model conditioned on privileged informat

safetyarxiv-cs-cl

1 Oct 2026

Who Verifies the Graph? Misspecification Attacks on Causal Action Verification for Language Agents

arXiv:2609.40027v1 Announce Type: new Abstract: Causal action verifiers gate an agent's state-changing tool calls by checking whether each proposed intervention is identifiable against a committed act

Why Do Conventional World Models Fail to Learn Cellular Automata?

arXiv:2609.39604v1 Announce Type: new Abstract: Although conventional world models - auto-regressive or diffusion models based on transformers or convolutional networks - may learn surface statistics

tutorialsarxiv-cs-ai

Research

X-Planner: Event-Structured Task Planning for Embodied Intelligence

arXiv:2609.25187v2 Announce Type: replace Abstract: Task planning bridges high-level instructions and executable behavior in long-horizon manipulation, yet modern Vision-Language-Action (VLA) systems

researcharxiv-cs-ai

ABC: Advantage-Based Control Variates for Reinforcement Learning with Verifiable Rewards

arXiv:2609.36058v1 Announce Type: new Abstract: Recent progress in reinforcement learning with verifiable rewards (RLVR) has highlighted the effectiveness of simple critic-free policy-gradient methods

safetyarxiv-cs-lg

Applications

Adversarial Defense in Cybersecurity: A Systematic Review of GANs for Threat Detection and Mitigation

arXiv:2509.20411v3 Announce Type: replace-cross Abstract: Machine learning-based cybersecurity systems are highly vulnerable to adversarial attacks, while Generative Adversarial Networks (GANs) act as

applicationsarxiv-cs-ai

Alignment Forecasting: Predicting Misalignment From Training Data

arXiv:2609.35805v1 Announce Type: cross Abstract: Training a language model on data with a narrow flaw can sometimes make the model broadly misaligned. Inspecting the data at face value often does not

ARCagent: An Adaptive Retrieval Calibration Agent for Clinical Question Answering

arXiv:2609.36392v1 Announce Type: new Abstract: In diseases where clinical guidelines are incomplete, contested, or mutually contradictory, knowledge completeness and dynamic conflict-aware synthesis

Auditable Long-Term Memory: A Deterministic Retrieval Chain Measured at 479/475 of 500 on LongMemEval-S

arXiv:2609.38021v1 Announce Type: cross Abstract: We evaluate an auditable long-term memory system on LongMemEval-S. Its retrieval chain uses hybrid candidate retrieval, cross-encoder reranking, cover

Billiger.de Products: A Bilingual Entity Matching Benchmark

arXiv:2609.37713v1 Announce Type: new Abstract: Existing product matching benchmarks primarily contain English-language product data and are often dominated by a single product category, such as elect

Agents

Bregman Consensus

arXiv:2609.35930v1 Announce Type: cross Abstract: Consider a community of agents who are seeking consensus on a set of parameters. The agents agree to use the same Bregman-type divergence to quantify

agentsarxiv-cs-lg

CLeaR: A Unified Framework for Resolving the Leakage-Degradation Dilemma in Style Transfer

arXiv:2609.38136v1 Announce Type: new Abstract: Style transfer aims to render target content in the style of a reference image, but existing methods often suffer from content leakage, where objects, l

safetyarxiv-cs-cv

Local Ai

Coherence-Aware Distributional Evaluation of Open-Ended Text Generation

arXiv:2609.34240v2 Announce Type: replace-cross Abstract: Existing open-ended generation metrics measure likelihood, lexical diversity, or distributional similarity in generic representation space, ye

local-aiarxiv-cs-ai

CompOrca: Corpus-Scale Compliance Labelling of Instruction-Tuning Data

arXiv:2609.37807v1 Announce Type: new Abstract: Studying how fine-tuning shapes refusal and noncompliance behaviour requires identifying training examples that refuse, evade or otherwise fail to fulfi

CruxBench: A Benchmark of Information Discovery

arXiv:2609.35879v1 Announce Type: cross Abstract: Benchmarks for large language models (LLMs) typically evaluate the accuracy of answers against fixed reference labels. But a central step in many comp

Distilling What Matters: Confidence-Aware Selective Distillation for Large Language Models

arXiv:2609.36734v1 Announce Type: new Abstract: Knowledge Distillation (KD) trains a smaller-capacity student model to imitate a larger-capacity teacher model by matching output distributions, implici

safetyarxiv-cs-cl

DraftTrace: A Multi-View Analytics Environment for AI-Integrated Writing

arXiv:2609.36544v1 Announce Type: cross Abstract: Generative AI has changed how students produce writing assignments. The final artifact is no longer sufficient to understand the process through which

tutorialsarxiv-cs-ai

Empower your agents with the Google Cloud CLI remote MCP server

model-releasesgoogle-cloud-ai

Today, we’re expanding our ecosystem of managed remote MCP servers by introducing the Google Cloud CLI remote MCP server in preview. Powered by the popular gcloud and bq (BigQuery) command-line tools,

Evidence-Guided Schema Normalization for Temporal Tabular Reasoning

arXiv:2512.00329v2 Announce Type: replace-cross Abstract: Temporal reasoning over evolving semi-structured tables poses a challenge to current QA systems. We propose an approach that recasts the task

Explore Broadly, Reason Sharply: Push Small Models toward the Frontier via Sampling

arXiv:2609.38104v1 Announce Type: new Abstract: Power-sharpened sampling is an inference-time alternative to reinforcement-learning (RL) post-training for enhancing reasoning in large language models

Fine-Tuning on Self-Generated and Reward-Weighted Data: Learning Dynamics, Convergence Rates, and Benefits of Off-Policyness

arXiv:2609.36945v1 Announce Type: new Abstract: We study the learning dynamics of fine-tuning a policy model on self-generated and reward-weighted data, with particular focus on a generalized version

safetyarxiv-cs-lg

Follow the Entities: A Corpus Map for Agentic Search

arXiv:2609.37226v1 Announce Type: cross Abstract: Answering questions and completing tasks over large document collections often requires connecting evidence spread across multiple documents, such as

From Retrieval to Reasoning: Agentic Mechanism Prediction from Cell Painting Profiles

arXiv:2609.36406v1 Announce Type: new Abstract: Cell Painting is a high-content morphological profiling assay widely used for phenotype-based biological inference, with mechanism of action (MOA) predi

GenLimitLib: A Formal Library for Language Generation in the Limit and AI-Assisted Mathematical Research

arXiv:2609.36663v1 Announce Type: new Abstract: We present GenLimitLib, a source-aligned Lean 4 library for language generation in the limit. Introduced by Kleinberg and Mullainathan at NeurIPS 2024,

tutorialsarxiv-cs-lg

GenomeOcean Anywhere: Private WebGPU Inference for Genome MoEs

arXiv:2609.35882v1 Announce Type: cross Abstract: Genome foundation models are most useful where sequences are generated, yet the largest models need datacenter accelerators and a place to send privat

Grounded Revision vs. Prior Injection: Probing Retrieval-Augmented Patent Claim Amendment

arXiv:2609.36550v1 Announce Type: new Abstract: Retrieval-augmented generation is widely used in professional writing, yet whether retrieval grounds revision or merely injects templates is rarely test

JudgeCast: Time Series Forecasting with Experience-Informed Covariate Judgements

arXiv:2609.36966v1 Announce Type: cross Abstract: Covariate effects vary across contexts and shift over time, requiring forecasters to assess how to use them for each forecasting context. As forecasti

tutorialsarxiv-cs-ai

KUPAS MASTER: Distilling the Tacit Expertise of Master Practitioners into Agent-Ready Experience Corpora

arXiv:2609.37673v1 Announce Type: new Abstract: Experienced professionals know more than just facts and conclusions. They know which cues matter, why a judgment is reasonable, and which action to take

Language Models Are 'Insecure' Reporters

arXiv:2609.36139v1 Announce Type: cross Abstract: As large language models are deployed in increasingly autonomous long-horizon tasks, manually auditing and verifying the actions, artifacts, and outpu

Large Language Models Exhibit Human-Like Bayesian Hypocrisy

arXiv:2609.35779v1 Announce Type: cross Abstract: Given recent achievements of large language models (LLMs), frontier models are expected to perform well on Bayesian reasoning tasks, at least as well

LatentSift: Policy-State Filtering for Token-Efficient Verification of Software Engineering Agents

arXiv:2609.36371v1 Announce Type: cross Abstract: Test-time scaling improves software engineering agents by generating multiple candidate trajectories and selecting the best one. Verifying and selecti

Applications

LAURA: Knowledge Distillation for Interpretable Ambiguous Clause Identification in Legal Contracts

arXiv:2609.36707v1 Announce Type: new Abstract: Legal contracts contain ambiguities that expose enterprises to financial and legal risks. Some ambiguities allow flexible interpretation without trigger

applicationsarxiv-cs-cl

Making Duplicate Reimbursement Unrepresentable: A Verified Ethereum E-Invoice System for Humans and AI Agents

arXiv:2609.37819v1 Announce Type: cross Abstract: Electronic invoices are replacing paper invoices worldwide, but today's centralized architectures leave three problems unsolved on the consumption sid

MatToolBench: Benchmarking Multimodal Agents in Real-World Materials Science Workflows

arXiv:2609.37053v1 Announce Type: new Abstract: Multimodal GUI agents have achieved impressive results on general software benchmarks, yet their ability to operate professional scientific software rem

Applications

Multi-Site Real-World Performance of Commercial AI for Pulmonary and Incidental Pulmonary Embolism Detection

arXiv:2609.37750v1 Announce Type: cross Abstract: Pulmonary embolism (PE) is a leading cause of cardiovascular mortality, yet the real-world performance of FDA-cleared AI detection models remains inco

applicationsarxiv-cs-ai

MyoCodec: A Streaming Neural Codec for Electromyography

arXiv:2609.36687v1 Announce Type: cross Abstract: Neural codecs encode continuous signals into compact sequences of discrete tokens, providing an interface for efficient transmission, storage, and tok

PR-OPD: Privileged Representation On-policy Self-Distillation for Agentic Reinforcement Learning

arXiv:2609.36642v1 Announce Type: new Abstract: Language-model agents are usually trained by reinforcement learning from one reward per episode, and privileged self-distillation enriches it by letting

safetyarxiv-cs-lg

ReCAP: Retrieval-Guided Capability Reuse for Multimodal Continual Instruction Tuning

model-releasesarxiv-cs-cv

arXiv:2609.37889v1 Announce Type: new Abstract: Multimodal continual instruction tuning (MCIT) aims to enable multimodal large language models to acquire new capabilities from sequential tasks while p

Reducing the Adaptation Gap Through Reachable Fisher Geometry

arXiv:2609.36329v1 Announce Type: new Abstract: Parameter-efficient fine-tuning (PEFT) determines not only how many parameters are trained, but also which local directions a model can move in, so simi

Replay the Curvature: Accurate and Scalable NVFP4 Quantization for Large Language Model Inference

arXiv:2609.36654v1 Announce Type: cross Abstract: Large language models make weight storage and memory traffic major inference costs, motivating low-precision formats that represent each weight with o

SAM Meets VLM: Parameter-Decoupled Full-Parameter Training for Unified Medical Reasoning and Segmentation

model-releasesarxiv-cs-cv

arXiv:2609.37283v1 Announce Type: new Abstract: Medical multimodal large language models (MLLMs) are increasingly expected not only to answer clinical questions, but also to localize the visual eviden

Sieve and Sage: Efficient Distraction Filtering for Reliable RALM Abstention

arXiv:2609.35794v1 Announce Type: new Abstract: Just as Socrates recognized the limits of his own knowledge, Retrieval-Augmented Language Models (RALMs) should learn to abstain when the retrieved evid

tutorialsarxiv-cs-cl

Single-turn emergency psychiatric triage across 15 frontier AI chatbots

arXiv:2604.25415v2 Announce Type: replace-cross Abstract: People increasingly turn to general-purpose AI chatbots for advice about emotional and mental health problems, but the ability of these system

Research

Soft Curriculum Learning for Optimizing Fresh and Generalized Recommendations

arXiv:2609.35783v1 Announce Type: cross Abstract: Large-scale recommender systems, particularly short-form video platforms, are often bottlenecked by massive popularity feedback loops. In such environ

researcharxiv-cs-lg

Systematic Multi-Agent Vision-and-Language Navigation: Formulation, Benchmark, and Method

arXiv:2609.35965v1 Announce Type: cross Abstract: Vision-and-Language Navigation (VLN) has largely focused on a single agent following a single instruction, yet many real-world applications require te

TabFM: A Zero-Shot Foundation Model for Tabular Data

arXiv:2609.37959v1 Announce Type: new Abstract: Tabular machine learning typically relies on per-dataset workflows, fitting tree ensembles or running AutoML searches from scratch for every task. We pr

VehicleArena: A Realistic Urban Environment for Multi-Agent Driving

arXiv:2609.35916v1 Announce Type: cross Abstract: Real-world embodied agents often pursue independent objectives within a shared physical environment, where their actions can alter the conditions face