# Entropy-KL Divergence-based Token Masking: A Novel Approach for Selective Fine-tuning of Large Language Models

arXiv: https://arxiv.org/abs/2605.29303 · HTML: https://mlsanity.com/paper/2605.29303/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 28 of its 43 references that are in the map.
- First posted on arXiv 28 May 2026.

## Builds on (most cited first, 28 of 28)

- [Training language models to follow instructions with human feedback](https://mlsanity.com/paper/2203.02155/) (2022)
- [Trust Region Policy Optimization](https://mlsanity.com/paper/1502.05477/) (2015)
- [HybridFlow: A Flexible and Efficient RLHF Framework](https://mlsanity.com/paper/2409.19256/) (2024)
- [Decoupled Weight Decay Regularization](https://mlsanity.com/paper/1711.05101/) (2017)
- [LlamaFactory: Unified Efficient Fine-Tuning of 100+ Language Models](https://mlsanity.com/paper/2403.13372/) (2024)
- [Qwen3 Embedding: Advancing Text Embedding and Reranking Through Foundation Models](https://mlsanity.com/paper/2506.05176/) (2025)
- [Tulu 3: Pushing Frontiers in Open Language Model Post-Training](https://mlsanity.com/paper/2411.15124/) (2024)
- [Large Language Monkeys: Scaling Inference Compute with Repeated Sampling](https://mlsanity.com/paper/2407.21787/) (2024)
- [Does Reinforcement Learning Really Incentivize Reasoning Capacity in LLMs Beyond the Base Model?](https://mlsanity.com/paper/2504.13837/) (2025)
- [SFT Memorizes, RL Generalizes: A Comparative Study of Foundation Model Post-training](https://mlsanity.com/paper/2501.17161/) (2025)
- [MathArena: Evaluating LLMs on Uncontaminated Math Competitions](https://mlsanity.com/paper/2505.23281/) (2025)
- [Learning to Reason under Off-Policy Guidance](https://mlsanity.com/paper/2504.14945/) (2025)
- [Reasoning with Exploration: An Entropy Perspective](https://mlsanity.com/paper/2506.14758/) (2025)
- [On Memorization of Large Language Models in Logical Reasoning](https://mlsanity.com/paper/2410.23123/) (2024)
- [On-Policy RL Meets Off-Policy Experts: Harmonizing Supervised Fine-Tuning and Reinforcement Learning via Dynamic Weighting](https://mlsanity.com/paper/2508.11408/) (2025)
- [SRFT: A Single-Stage Method with Supervised and Reinforcement Fine-Tuning for Reasoning](https://mlsanity.com/paper/2506.19767/) (2025)
- [AceReason-Nemotron 1.1: Advancing Math and Code Reasoning through SFT and RL Synergy](https://mlsanity.com/paper/2506.13284/) (2025)
- [Nemotron-Research-Tool-N1: Exploring Tool-Using Language Models with Reinforced Reasoning](https://mlsanity.com/paper/2505.00024/) (2025)
- [UFT: Unifying Supervised and Reinforcement Fine-Tuning](https://mlsanity.com/paper/2505.16984/) (2025)
- [olmOCR 2: Unit Test Rewards for Document OCR](https://mlsanity.com/paper/2510.19817/) (2025)
- [Supervised Fine Tuning on Curated Data is Reinforcement Learning (and can be improved)](https://mlsanity.com/paper/2507.12856/) (2025)
- [AlphaMaze: Enhancing Large Language Models' Spatial Intelligence via GRPO](https://mlsanity.com/paper/2502.14669/) (2025)
- [From Single to Multi-Granularity: Toward Long-Term Memory Association and Selection of Conversational Agents](https://mlsanity.com/paper/2505.19549/) (2025)
- [Harnessing Large Language Models for Knowledge Graph Question Answering via Adaptive Multi-Aspect Retrieval-Augmentation](https://mlsanity.com/paper/2412.18537/) (2024)
- [Rethinking Conventional Wisdom in Machine Learning: From Generalization to Scaling](https://mlsanity.com/paper/2409.15156/) (2024)
- [MedGR2: Breaking the Data Barrier for Medical Reasoning via Generative Reward Learning](https://mlsanity.com/paper/2508.20549/) (2025)
- [AMFT: Aligning LLM Reasoners by Meta-Learning the Optimal Imitation-Exploration Balance](https://mlsanity.com/paper/2508.06944/) (2025)
- [Making Qwen3 Think in Korean with Reinforcement Learning](https://mlsanity.com/paper/2508.10355/) (2025)

## Built on by (most cited first, 1 of 1)

- [Rethinking Token Reweighting for SFT: Suppress, Reverse, and Extrapolate Learned Features](https://mlsanity.com/paper/2609.33463/) (2026)

## Abstract (arXiv)

Supervised fine-tuning (SFT) followed by reinforcement learning (RL) has become a standard post-training paradigm for large language models. This paradigm provides a cold-start for RL exploration, avoiding the inefficiency of pure RL where on-policy sampling yields insufficient positive samples. However, in practice, existing approaches often use a small amount of data for SFT initialization compared to the RL phase, which can cause the model to fit the limited samples and shift away from its pre-trained distribution. This distribution shift impedes the model's ability to effectively explore during subsequent RL training. To address this challenge, we propose that in low-data regimes, SFT should prioritize activating task-relevant capabilities rather than memorizing specific content. Along this line, we propose EKSFT (Entropy-KL Selective Fine-Tuning), which selectively masks tokens that exhibit either high entropy or high KL divergence from a reference model. By excluding these high-uncertainty, distribution-shifting tokens from imitation, EKSFT injects task-specific knowledge while preserving the integrity of the model's pre-trained distribution. Empirical evaluations on mathematical reasoning benchmarks demonstrate that EKSFT consistently outperforms standard SFT. Further RL fine-tuning from the EKSFT model yields consistently better post-RL performance, indicating improved exploration for the RL stage. Our codes and datasets are available at https://github.com/MINE-USTC/EKSFT.