This paper analyzes the limitations of Counterfactual Knowledge Training (CFT) for LLM unlearning, identifying knowledge conflict and hallucination spillover as major pitfalls that hinder its effectiv…

This paper analyzes the internal decision-making process of large language models by tracking how the answer score changes across multiple internal computational steps (trajectories), finding that mod…

The paper decomposes LLM debate convergence into three mechanisms (instability, conformity, persuasion) and finds that much observed convergence is harmful social compliance rather than genuine reason…

The paper proposes Resonant Context Anchoring (RCA), a lightweight, training-free method that enhances factual faithfulness in LLMs by dynamically amplifying the signal of external context evidence du…

The paper identifies a failure mode called unfaithful capitulation (UC), where reasoning models maintain a correct internal thought process (chain-of-thought) but output an incorrect final answer when…

The paper introduces COMET, a novel PLS-SVD framework, to analyze the audio-text modality gap in CLAP models, showing that shared concepts are captured by a small subset of axes, and proposes a spectr…

The paper proposes Reasoning-Conditioned Direct Preference Optimization (RC-DPO) to effectively mitigate hallucinations in multimodal large reasoning models by explicitly conditioning the preference o…

PolySpeech-100 introduces a massive, multi-lingual benchmark covering 110 linguistic variants to rigorously test Speech-LLMs, demonstrating that open-source models struggle with low-resource languages…

The paper introduces three new Korean speech benchmarks (KVoiceBench, KOpenAudioBench, and KMMAU) to evaluate SpeechLMs, demonstrating that English-centric evaluation fails to capture performance gaps…

The paper introduces Responsible Contrastive Soft Prompting (RCSP), a parameter-efficient method using soft prompts to improve LLM reliability by simultaneously suppressing hallucinations, encouraging…

COFT is a training-free decoding method that significantly reduces societal biases in large language model chain-of-thought reasoning by applying token-level fairness control at decode time.

This paper systematically measures and explains how sequential model defenses can conflict, finding that 38.9% of ordered defense sequences cause measurable risk exacerbation due to anti-aligned param…