AI Paper Atlas

Research map / Large language models / Alignment and preferences

TokenRatio: Principled Token-Level Preference Optimization via Ratio Matching

Truong Nguyen, Tien-Phat Nguyen, Linh Ngo Van, Duy Minh Ho Nguyen, Khoa D Doan, Trung Le

ICML 2026 · Large language models › Alignment and preferences · approach: DPO design choices · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
token-level Bregman ratio matching (TBPO-Q) instead of sequence-level DPODPO, Mistral 7B v0.1 · Open LLM Leaderboard averageaverage score: 59.30 -> 60.95Improved
token-level Bregman ratio matching (TBPO-A) instead of sequence-level DPODPO, Llama 3 8B · Open LLM Leaderboard averageaverage score: 66.24 -> 70.50Improved
TBPO-Q instead of DPODPO, Mistral 7B v0.1 · GSM8Kaccuracy: 34.87 -> 39.34Improved
TBPO-Q instead of DPODPO, Mistral 7B v0.1 · TruthfulQAaccuracy: 42.30 -> 44.47Improved
TBPO-A instead of TDPOTDPO, Llama 3 8B · ARCaccuracy: 59.04 -> 64.08Improved
learned baseline head (Q) versus KL-based baseline (A) for the state weightTBPO · Open LLM Leaderboard averageaverage score: 60.95 vs 60.73 (Mistral); 70.38 vs 70.50 (Llama 3)No clear effect
TBPO instead of baselines for pairwise judgingTDPO, DPO, TIS-DPO, BPO · MT-Benchpairwise win rateImproved
TBPO instead of baselinesDPO and token-level baselines, Llama 3 8B · Anthropic HH-RLHFLC win rateImproved
TBPO instead of DPO-style methodsDPO, TDPO, BPO · UltraFeedback generation diversitypredictive entropy, Distinct-1, Self-BLEUImproved
TIS-DPO token-level importance samplingDPO · UltraFeedback generation diversitypredictive entropy, Self-BLEUHurt
TDPO token-wise decomposition instead of DPODPO, Mistral 7B v0.1 · GSM8Kaccuracy: 34.87 -> 28.65Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy