Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Qing, Liu, Wu, Ou, Du, Yi
Format: Preprint
Veröffentlicht: 2026
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2606.01635
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866918534532038656
author Qing, Liu
Wu, Ou
Du, Yi
author_facet Qing, Liu
Wu, Ou
Du, Yi
contents Token selection is pivotal for effective LLM post-training. However, existing methods mostly rely on local heuristics and rarely formulate token selection as a principled valuation of individual response tokens. We introduce $\textbf{AlphaToken}$, a response token valuation framework that decouples valuation into $\textbf{adaptation}$ (promoting target-task learning) and $\textbf{stability}$ (preserving pre-trained capabilities), and makes each objective $\textbf{path-aware}$ by combining the direct-path signal from local token gradients with the downstream causal-path signal in autoregressive generation. Since retention data are typically unavailable, AlphaToken approximates stability via a $\textbf{Fisher-drift proxy}$ anchored at the pre-trained reference model. For efficient computation, we extend Ghost Dot-Product to token-level valuation. AlphaToken masks low-value response tokens during fine-tuning and preference optimization, concentrating training signals on more valuable positions. Experiments show that AlphaToken improves post-training performance and mitigates catastrophic forgetting.
format Preprint
id arxiv_https___arxiv_org_abs_2606_01635
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle AlphaToken: Decoupling Adaptation and Stability for Path-Aware Response Token Valuation in LLM Post-Training
Qing, Liu
Wu, Ou
Du, Yi
Computation and Language
Artificial Intelligence
Token selection is pivotal for effective LLM post-training. However, existing methods mostly rely on local heuristics and rarely formulate token selection as a principled valuation of individual response tokens. We introduce $\textbf{AlphaToken}$, a response token valuation framework that decouples valuation into $\textbf{adaptation}$ (promoting target-task learning) and $\textbf{stability}$ (preserving pre-trained capabilities), and makes each objective $\textbf{path-aware}$ by combining the direct-path signal from local token gradients with the downstream causal-path signal in autoregressive generation. Since retention data are typically unavailable, AlphaToken approximates stability via a $\textbf{Fisher-drift proxy}$ anchored at the pre-trained reference model. For efficient computation, we extend Ghost Dot-Product to token-level valuation. AlphaToken masks low-value response tokens during fine-tuning and preference optimization, concentrating training signals on more valuable positions. Experiments show that AlphaToken improves post-training performance and mitigates catastrophic forgetting.
title AlphaToken: Decoupling Adaptation and Stability for Path-Aware Response Token Valuation in LLM Post-Training
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2606.01635