#reasoning

try —

11 papers match

cs.CL2026

Can Agents Deceive? Evaluating Reasoning and Deception in ParliamentBench using a Social Deduction Game

Niklas Bauer, Lars Benedikt Kaesberg, Akiko Aizawa +3

The paper introduces ParliamentBench, an open-source benchmark based on the Secret Hitler game, to evaluate large language model agents on deception, persuasion, and reasoning unde…

#large language models#deception#social deduction games#agent evaluation
cs.CV2026

FAS-R1: A Unified Multi-Task MLLM for Reasoning Face Anti-Spoofing

Hongyang Wang, Yichen Shi, Hongrui Li +3

The paper introduces FAS-R1, a two‑stage multimodal large language model that simultaneously classifies face authenticity, identifies attack types, and localizes spoof regions, usi…

#face anti-spoofing#multimodal large language models#reasoning#attack type recognition
cs.CV2026

Understanding Knowledge Transfer Mechanism in Heterogeneous MLLM Fusion: A Simple Linear Approach

Yinghao Hou, Jiahe Fan, Yuanhao Pu +2

The paper introduces a simple linear probe called Cross-Scale Directional Parameter Injection (CDPI) to study how knowledge transfers when heterogeneous multimodal large language m…

#multimodal large language models#model fusion#knowledge transfer#linear probing
cs.LG2026

Learning Mechanistic Reasoning for Chemical Reactions with Large Language Models

Xingyu Dang, Haocheng Tang, Junmei Wang +1

The paper introduces a large-scale dataset and a benchmark (FukuyamaBench) for teaching large language models to reason step‑by‑step about chemical reaction mechanisms, and shows t…

#chemical reaction mechanisms#large language models#reasoning#benchmark
cs.LG2026

On-Policy Delta Distillation

Byeongho Heo, Jaehui Hwang, Sangdoo Yun +1

The paper proposes On-Policy Delta Distillation (OPD²), a new on‑policy distillation method that uses a delta signal—the difference between a teacher LLM and its pre‑tuned base mod…

#on-policy distillation#reinforcement learning#large language models#reasoning
cs.AI2026

Policy of Thoughts: Scaling Test-Time Training for LLM Reasoning via Online Policy Evolution

Zhengbo Jiao, Hongyu Xian, Qinglong Wang +5

The paper introduces Policy of Thoughts (PoT), a test‑time training framework that continuously updates a lightweight LoRA adapter using online policy optimization to improve large…

#large language models#test-time training#online policy optimization#reasoning