Functional interpretation is introduced, which characterizes an SAE feature as a mapping from its activating input semantics to its output effects under intervention, and Dual-End Agentic Feature Interpretation (DAFI) is presented, an agent that actively gathers evidence and refines input-side, output-side, and functio...
De-Wen Liu, Zi-Xu Li, Jonathan Pan et al.· 0 citations
SAEVerbalizer is introduced, a framework that injects SAE decoder directions into an LLM's representations and fine-tunes the LLM's downstream layers to generate natural-language explanations of the injected features.
Weihang Meng, Hong-Zhu Guo, Yi Jing et al.· 0 citations
Results support a qualified internalized-search reading: under the recipe the authors test, much of the measured RL gain corresponds to a change in sampling efficiency toward operating points the base model can already reach under search.
Wen-He Sun, Cun-Xiang Wang, Zi-Jun Yao et al.· 0 citations
Tail subtraction is introduced, which removes shared prompt and continuation semantics from boundary states and yields cleaner, more stable steering signals, and suggests that steering depends on representations of what the model is about to do, not merely on what has already appeared.
Jiaran Ye, Lingxu Ran, Zijun Yao et al.· arXiv.org· 2 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
