Same Bytes, Different Authority: Reserved-Token Representations in Chat-Template Prompt Injection Paper • 2609.35932 • Published Sep 28 • 7
Keyword Harnesses Fail Open: A Cheap Diagnostic Ladder for Tool-Use Claims in Small Language Models Paper • 2610.02142 • Published Oct 1 • 4
SpecFold: Folding Multi-Branch Redundancy for Faster Speculative Decoding in Diffusion Language Models Paper • 2610.04875 • Published Oct 6 • 6
MARGIN: Runtime Confidence Calibration for Multi-Agent Foundation Model Coordination Paper • 2605.22949 • Published Oct 8 • 4
AutoResearch at Production Scale: Failure Modes and a Multi-Agent Framework Paper • 2609.30541 • Published Sep 24 • 5
Agent Plasticity: Measuring Self-Improvement Through Experience Paper • 2610.08902 • Published Oct 6 • 8
CheckerBench: Can Long-Horizon Agents Synthesize Static-Analysis Checkers? Paper • 2610.07557 • Published Oct 6 • 12
Equal CER, Unequal Damage: Severity-Weighted Character Error Rate for OCR Paper • 2610.20461 • Published Oct 14 • 4
DoubtLess: Training-Free Contrastive Decoding for Efficient Reasoning Paper • 2610.18862 • Published Oct 13 • 7
Probable but Not Valid: A Psychometric Audit of Decision Models as Synthetic Survey Respondents Paper • 2610.14912 • Published Oct 10 • 4
Words to World: Benchmarking LLMs in Interactive Physical Environments Paper • 2610.15228 • Published Oct 10 • 5
Scaling Verifiable Environments for Long-horizon Work Agents Paper • 2610.04906 • Published Oct 8 • 7
Training Reasoning Models to Generate Diverse Scientific Ideas Paper • 2610.22324 • Published Oct 15 • 7
MultiAgent-OPD: End-to-End Training of Hierarchical Multi-Agent Systems with On-Policy Distillation Paper • 2610.23234 • Published 27 days ago • 12
EnvLunch: The Environment as a Free Lunch for Learning to Interact Paper • 2610.33817 • Published 23 days ago • 5
ThinkBit: Search Codewords, Recalibrate Scales, Select by Reasoning Paper • 2610.34886 • Published 21 days ago • 4