Search: “gpt”

Can agents whose sole duty is to keep humans alive be hijacked to kill? Delayed multi-turn derived-action trap attacks in simulated high-stakes environments

arXiv:2610.15153v1 Announce Type: new Abstract: We ask whether a tool-using agent whose sole duty is to keep humans alive can be induced, by delayed prompt injection, to kill. The keep-alive duty is i

Human

Content type

All Blog X Post Paper YouTube Reddit GitHub

Clear filters

ChemArgus: Automated Scientific Diagnostics for Lab-Grade Multimodal Chemistry Reasoning: A Frontier Benchmark

arXiv:2610.20012v1 Announce Type: new Abstract: Chemistry evaluation has moved past multiple-choice probes toward open-ended expert problems, yet the benchmarks that define the current frontier still

Does a Governed Confidence Score Know When a Text-to-SQL Answer Is Right? An Exploratory Study on BIRD and a Pre-registered Confirmation on Spider

arXiv:2610.17256v1 Announce Type: new Abstract: Conversational analytics lets people ask questions of records spread across many systems and get answers in plain language; it is useful only if each an

Evaluating Large Language Models for Multilingual Abstractive Summarization: A Comprehensive Study of English and Bengali

arXiv:2610.19003v1 Announce Type: new Abstract: Abstractive summarization remains challenging across languages, where linguistic quality and factual consistency may diverge, particularly in low-resour

Evaluating LLMs as Agentic Workflow Executors in Real-World Customer Support Skills

arXiv:2610.14346v2 Announce Type: replace Abstract: Enterprise customer-support agents increasingly rely on Large Language Models (LLMs) to execute bounded, tool-using workflows. We study this problem

From Unreliable to Usable: Decomposing Failure Modes and Confidence Signals in LLM Entity Extraction - Evidence from Cyber Threat Intelligence

arXiv:2610.20417v1 Announce Type: new Abstract: Audits of large language model (LLM) extraction from full-length documents report severe unreliability, leaving two questions open: what the failures ac

Gray-Box Hallucination Detection in LLMs with Token Identities

arXiv:2610.19996v1 Announce Type: new Abstract: Large language models often produce fluent but factually incorrect output. Gray-box detectors identify such hallucinations from the token-probability st

Applications

Have You Been Convinced? A Neuro-symbolic Argumentation Reasoning Approach to Detecting Persuasiveness in ChangeMyView Discussions

arXiv:2610.15091v1 Announce Type: cross Abstract: Argument structures represent how individual arguments interact within a discourse. However, leveraging these structures to infer meaningful knowledge

applicationsarxiv-cs-ai

High-Probability Regret Bounds for Feel-Good Thompson Sampling

arXiv:2610.19513v1 Announce Type: cross Abstract: Feel-good Thompson sampling (FGTS) augments the posterior of vanilla Thompson sampling with an exploration term that favors reward functions with high

Humans vs. Machines: A Comparison With AI Agents in Software Reverse Engineering

arXiv:2610.14960v1 Announce Type: cross Abstract: Can an autonomous AI agent reverse engineer a binary as well as a skilled human analyst? Do they adopt the same strategy, and can agents scale to real

Instructions Are Evidence, Not Objectives: Measuring Outcome Alignment of Agents

arXiv:2610.17757v1 Announce Type: new Abstract: Large language model agents can follow complex human instructions to solve real-world tasks. In practice, however, initial instructions are often incomp

Safety

Learning Perception to Action Rules through Reinforcement Learning for Theory-of-Mind Reasoning

arXiv:2610.19796v1 Announce Type: cross Abstract: In Theory-of-Mind (ToM) reasoning, large language models may correctly infer the agent's mental state but fail to use that inference when predicting o

safetyarxiv-cs-cl

LED-BENCH: A Benchmark for Large-format Engineering Drawing Understanding

arXiv:2610.19571v1 Announce Type: new Abstract: Real-world engineering drawings are often large-format and information-dense. Answering practical engineering questions therefore requires both fine-gra

LifeSearch: Benchmarking Deep Research Agents for Real-World Life and Productivity

arXiv:2610.18092v1 Announce Type: new Abstract: Deep research (DR) agents address open-ended information needs through multi-step reasoning, long-horizon web search, and evidence synthesis. However, e

LLM Alignment--Utility Asymmetry under Semantic-Preserving Transformations

arXiv:2609.32717v2 Announce Type: replace-cross Abstract: Large Language Model (LLM) alignment is intended to ensure that models remain helpful and safe, but its stability under input distributional s

Measuring Epistemic Humility in Multimodal Large Language Models

arXiv:2509.09658v3 Announce Type: replace Abstract: Hallucinations in multimodal large language models (MLLMs), where a model generates content inconsistent with the input image, pose significant risk

MolHop: Scaffold-Hopping Molecular Design with Agentic Reinforcement Learning

arXiv:2610.19920v1 Announce Type: cross Abstract: Molecular design is often formulated as an optimization problem over existing molecular spaces. However, molecular optimization and the discovery of s

PreciseCoder: Faithful and Generalizable Debugging Beyond Unit-Test Rewards

arXiv:2610.18755v1 Announce Type: cross Abstract: Reinforcement learning (RL) from unit tests is the common recipe for post-training code models, but a unit test only asks whether the final program pa

Probable but Not Valid: A Psychometric Audit of Decision Models as Synthetic Survey Respondents

arXiv:2610.14912v1 Announce Type: new Abstract: Language models are increasingly used to create synthetic survey samples, also called silicon samples. Decision models are a new class of language model

PsyMemBehav: Psychology-Informed and Memory-Grounded Two-Level Identity-to-Action Reasoning for Social-Media Behavior Simulation

arXiv:2610.14570v1 Announce Type: new Abstract: Large language models (LLMs) support social simulation, but faithfully reproducing real users requires grounding both structured platform decisions and

RAN-JEPA: An Action-Conditioned World Model for Energy-efficient Mobile Network Control

arXiv:2610.17584v1 Announce Type: new Abstract: Operational mobile networks are rich in observations but poor in interventions: they continuously generate large volumes of passive telemetry, while con

researcharxiv-cs-ai

Relevant to the comprehension challenge I posed in 2014 and my bet with Miles Brundage.

model-releasesgary-marcus--x

Relevant to the comprehension challenge I posed in 2014 and my bet with Miles Brundage. We edited Mark Twain's Huckleberry Finn and fed it to 9 AI models to summarize a chapter, twice each. Only Claud

REPRO: Reflective Evolutionary Prompt Optimizer for Automated LLM Red-Teaming

arXiv:2610.15513v1 Announce Type: cross Abstract: Safety-aligned large language models (LLMs) remain vulnerable to adversarial prompts that bypass their safety training. Existing automated red-teaming

RoTA-Bench: Evaluating 3D Spatial Reasoning through Interactive Projection Alignment

arXiv:2610.17665v2 Announce Type: new Abstract: Can vision-language models turn 2D observations into 3D understanding that supports precise action? This requires inferring object geometry, anticipatin

Smart Alone, Blind Together: Why Accuracy Fails to CertifyTheory of Mind in Multi-Agent Teams

arXiv:2610.19051v1 Announce Type: cross Abstract: Coordinators that act on what they infer about teammates are usually certified by the accuracy of those inferences. We argue that accuracy is the wron

Specification-Based Testing of Large Language Models

arXiv:2610.14443v1 Announce Type: cross Abstract: Model cards, system cards, and safety reports increasingly document LLM safety behaviors, yet safety evaluations typically start from predefined bench

TextualRL: Textual Reinforcement Learning for Context Management

arXiv:2610.18216v1 Announce Type: cross Abstract: LLM agents can adapt to task feedback by revising the textual context that guides their behavior, without changing their model parameters. However, tu

Training LLMs Beyond Next Token Prediction - Filling the Mutual Information Gap

arXiv:2511.00198v2 Announce Type: replace-cross Abstract: Autoregressive language models are usually trained with next-token prediction under a fixed left-to-right target order. This paper studies tar

Uncovering Robustness Weaknesses in LLM Agents via Persona Evolution with Task-Integrity Constraints

arXiv:2610.16155v1 Announce Type: new Abstract: Interactive-agent evaluations commonly rely on standardized LLM-based user simulators that exhibit broadly cooperative behavior, potentially overlooking

Tutorials

What Do Foundation Models for Market Microstructure Learn? A Controlled Study with Synthetic Limit Order Books

arXiv:2610.20104v1 Announce Type: new Abstract: Autoregressive transformers trained on limit order book (LOB) message streams are increasingly proposed as foundation models for market microstructure.

tutorialsarxiv-cs-lg

A Refresh Interval Does Not Fully Specify a Refresh Schedule: Learning-Rate-Dependent Placement in Kronecker-Preconditioned Language Model Training

arXiv:2610.18695v1 Announce Type: new Abstract: Kronecker-preconditioned optimizers recompute a cached curvature eigendecomposition only periodically, and implementations and adaptive methods control

Local Ai

Anomaly-Guided Observation and Two-Level Semantic Feedback for Training-Free Video Anomaly Analysis

arXiv:2610.16324v1 Announce Type: new Abstract: Training-free language-based video anomaly detection (VAD) uses frozen vision--language models (VLMs) and large language models (LLMs) to localize abnor

local-aiarxiv-cs-cv

AutoEnvScaling: Automating the Data Flywheel with Terminal Agents

arXiv:2610.14163v1 Announce Type: cross Abstract: Terminal agents solve tasks in software engineering and scientific research by executing commands in a terminal. Training them via reinforcement learn

Before Attention Sinks Form: Hidden-State Geometry Reveals and Modulates Their Formation

arXiv:2610.16745v1 Announce Type: new Abstract: Attention sinks are usually studied after they have formed. We ask whether a measurable hidden-state change appears beforehand and whether changing that

BSIPipe: Bidirectional Sequence-Level Interleaved Pipeline Parallelism for Long-Sequence Large Language Model Training

arXiv:2610.14935v1 Announce Type: new Abstract: Pipeline parallelism has become a critical technique for training large language models. To address the throughput limitation of existing sequence-level

Building Multilingual Datasets for Predicting Mental Health Severity through LLMs: Prospects and Challenges

arXiv:2409.17397v4 Announce Type: replace-cross Abstract: Large Language Models (LLMs) are increasingly being integrated into various medical fields, including mental health support systems. However,

Code-Interactive Visual Reasoning for Financial Candlestick Understanding

arXiv:2610.16587v1 Announce Type: new Abstract: Financial candlestick understanding requires locating visual elements, aligning them across trading dates and panels, and combining observations through

Conditional Multi-Event Temporal Grounding in Long-Form Video

arXiv:2606.15320v2 Announce Type: replace Abstract: Multimodal large language models have made rapid progress in video temporal grounding, yet real-world applications routinely require localizing ever

Diagnose-to-Generate: Eliminating Structural Distortions in Flow Models via Patch-Level Grounded Alignment

arXiv:2610.17882v1 Announce Type: new Abstract: Flow-based image generators achieve strong semantic synthesis but still produce localized structural distortions, such as malformed hands and limbs. Exi

Discover AI: A Learning Guide - 51 hands-on labs for learning how AI actually works.

model-releasesr-localllama

This is an offline app with 51 hands-on labs for learning how AI actually works, from neurons to agents and more... **A bit of insight** - 51 labs in 6 groups, from the basics (what a neuron is, gradi

Does System One Transfer? Calibrated Typed-Decision Models, Label Budgets, and Translation for Indonesian Regional Languages

arXiv:2610.14112v1 Announce Type: cross Abstract: Calibrated typed-decision ('System One') models return a label and a probability in a single forward pass at a small fraction of the cost of large lan

Safety

EgoMatrix: Physically Faithful Dexterous Robot Demonstrations from Egocentric Human Videos

arXiv:2610.18822v1 Announce Type: new Abstract: Egocentric human videos offer a scalable alternative to robot teleoperation for collecting dexterous manipulation data. However, two challenges limit th

safetyarxiv-cs-ro

Evaluating the Causal Faithfulness of SHAP as a Candidate Attribution Method for SAE Features

arXiv:2610.17898v1 Announce Type: new Abstract: Sparse autoencoders decompose language model representations into interpretable features, but feature interpretation in this literature remains predomin

Evidence-First Agentic Verification

arXiv:2610.17724v1 Announce Type: cross Abstract: Long-horizon, semi-verifiable agentic tasks often rely on rubric-based evaluation, but the evidence needed to assess these tasks can be spread across

GIFT-Eval-CTX: A Benchmark For Contextual Forecasting Model Evaluation

arXiv:2610.17794v1 Announce Type: new Abstract: Contextual forecasting seeks to improve time series forecasting by incorporating forward-looking information, such as planned events, external condition

HarnessTTT: Learn to Optimize Harness at Test-Time for Scientific Discovery

arXiv:2610.17370v1 Announce Type: new Abstract: Scientific discovery with language models requires finding the best valid solution within a finite computational budget. We study how feedback from this

I built BitRun: A standalone runner for 1-bit LLMs (100–127 t/s on Apple Silicon & 40–45 t/s on PC CPUs with zero compile)

model-releasesr-localllama

I built BitRun because I wanted to run 1-bit ternary LLMs on ordinary laptops without melting my battery, spinning up a dedicated GPU, or wrestling with complex build environments. Right now, BitRun g

LLM Agents for Automated Post-Training via Structured Recipe Optimization

arXiv:2610.18572v1 Announce Type: cross Abstract: Automated post-training transforms pre-trained large language models (LLMs) into models capable of handling diverse domains by automatically construct