5 papers
STR: Supervised Transcoder Replacement for Reducing Steering Side Effects
Haonan Yu, Junhao Liu, Zhenyu Yan +2
Model steering can strengthen a target behavior while degrading other useful behaviors. We introduce Supervised Transcoder Replacement (STR) to reduce these side effects for existi…
Guiding LLM-based Loop Invariant Synthesis via Feedback on Local Reasoning Errors
Tianchi Li, Zhenyu Yan, Junhao Liu +2
We propose a novel framework that provides constructive feedback to an LLM in the "guess-and-check" paradigm by formally verifying its own thinking process and detecting local reas…
WASD: Locating Critical Neurons as Sufficient Conditions for Explaining and Controlling LLM Behavior
Haonan Yu, Junhao Liu, Zhenyu Yan +2
Precise behavioral control of large language models (LLMs) is critical for complex applications. However, existing methods often incur high training costs, lack natural language co…
Focus-LIME: Surgical Interpretation of Long-Context Large Language Models via Proxy-Based Neighborhood Selection
Junhao Liu, Haonan Yu, Zhenyu Yan +1
As Large Language Models (LLMs) scale to handle massive context windows, achieving surgical feature-level interpretation is essential for high-stakes tasks like legal auditing and…
Revitalizing Black-Box Interpretability: Actionable Interpretability for LLMs via Proxy Models
Junhao Liu, Haonan Yu, Zhenyu Yan +1
Post-hoc explanations provide transparency and are essential for guiding model optimization, such as prompt engineering and data sanitation. However, applying model-agnostic techni…