@inproceedings{liu-etal-2026-alphatoken,
title = "{A}lpha{T}oken: Decoupling Adaptation and Stability for Path-Aware Response Token Valuation in {LLM} Post-Training",
author = "Liu, Qing and
Wu, Ou and
Du, Yi",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-main.866/",
pages = "20028--20054",
ISBN = "979-8-89176-459-0",
abstract = "Token selection is pivotal for effective LLM post-training. However, existing methods mostly rely on local token-level heuristics, while response-token selection in post-training lacks an optimization-guided valuation proxy. 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. Code at https://github.com/luciabcd/Alphatoken."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="liu-etal-2026-alphatoken">
<titleInfo>
<title>AlphaToken: Decoupling Adaptation and Stability for Path-Aware Response Token Valuation in LLM Post-Training</title>
</titleInfo>
<name type="personal">
<namePart type="given">Qing</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ou</namePart>
<namePart type="family">Wu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yi</namePart>
<namePart type="family">Du</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hung-Yi</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sunipa</namePart>
<namePart type="family">Dev</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jessy</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Desmond</namePart>
<namePart type="family">Elliott</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-459-0</identifier>
</relatedItem>
<abstract>Token selection is pivotal for effective LLM post-training. However, existing methods mostly rely on local token-level heuristics, while response-token selection in post-training lacks an optimization-guided valuation proxy. We introduce AlphaToken, a response token valuation framework that decouples valuation into adaptation (promoting target-task learning) and stability (preserving pre-trained capabilities), and makes each objective 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 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. Code at https://github.com/luciabcd/Alphatoken.</abstract>
<identifier type="citekey">liu-etal-2026-alphatoken</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.866/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>20028</start>
<end>20054</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T AlphaToken: Decoupling Adaptation and Stability for Path-Aware Response Token Valuation in LLM Post-Training
%A Liu, Qing
%A Wu, Ou
%A Du, Yi
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-459-0
%F liu-etal-2026-alphatoken
%X Token selection is pivotal for effective LLM post-training. However, existing methods mostly rely on local token-level heuristics, while response-token selection in post-training lacks an optimization-guided valuation proxy. We introduce AlphaToken, a response token valuation framework that decouples valuation into adaptation (promoting target-task learning) and stability (preserving pre-trained capabilities), and makes each objective 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 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. Code at https://github.com/luciabcd/Alphatoken.
%U https://aclanthology.org/2026.emnlp-main.866/
%P 20028-20054
Markdown (Informal)
[AlphaToken: Decoupling Adaptation and Stability for Path-Aware Response Token Valuation in LLM Post-Training](https://aclanthology.org/2026.emnlp-main.866/) (Liu et al., EMNLP 2026)
ACL