18 citations · 27 across the 32 of their papers we have counts for
6 papers · 1 filter
LaVGuard: Selective Decoding for Contextual Value Drift
Jiale Dai, Liuxian Ma, Hongcan Deng +5
A content-safe response can still promote priorities that conflict with its task: an assistant may demand compliance when asked to preserve autonomy. We study contextual value drif…
VISTA: Value-Informed Event Appraisal for Multimodal Emotion Conflict
Jiale Dai, Liuxian Ma, Xiaoke Niu +5
Conflicting emotional cues can be individually valid: a subdued voice may reflect a blocked goal while a smile satisfies a social obligation. Their interpretation depends on what t…
HEAL: Hindsight Entropy-Assisted Learning for Reasoning Distillation
Wenjing Zhang, Jiangze Yan, Jieyun Huang +7
Distilling reasoning capabilities from Large Reasoning Models (LRMs) into smaller models is typically constrained by the limitations of rejection sampling. Standard methods treat t…
Quantifying the Capability Boundary of DeepSeek Models: An Application-Driven Performance Analysis
Kaikai Zhao, Zhaoxiang Liu, Xuejiao Lei +12
DeepSeek-R1, known for its low training cost and exceptional reasoning capabilities, has achieved state-of-the-art performance on various benchmarks. However, detailed evaluations…
GlitchMiner: Mining Glitch Tokens in Large Language Models via Gradient-based Discrete Optimization
Zihui Wu, Haichang Gao, Ping Wang +3
Glitch tokens, inputs that trigger unpredictable or anomalous behavior in Large Language Models (LLMs), pose significant challenges to model reliability and safety. Existing detect…
Piculet: Specialized Models-Guided Hallucination Decrease for MultiModal Large Language Models
Kohou Wang, Xiang Liu, Zhaoxiang Liu +2
Multimodal Large Language Models (MLLMs) have made significant progress in bridging the gap between visual and language modalities. However, hallucinations in MLLMs, where the gene…