Cover Image

ProFit: Leveraging High-Value Signals in SFT via Probability-Guided Token Selection

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Liu, Tao, Wu, Taiqiang, Yang, Runming, Sun, Shaoning, Wang, Junjie, Yang, Yujiu
Format:	Preprint
Published:	2026
Subjects:	Computation and Language Artificial Intelligence
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

LoCa: Logit Calibration for Knowledge Distillation
by: Yang, Runming, et al.
Published: (2024)

Shadow-FT: Tuning Instruct Model via Training on Paired Base Model
by: Wu, Taiqiang, et al.
Published: (2025)

Revisiting Model Interpolation for Efficient Reasoning
by: Wu, Taiqiang, et al.
Published: (2025)

Timber: Training-free Instruct Model Refining with Base via Effective Rank
by: Wu, Taiqiang, et al.
Published: (2025)

LLM-NEO: Parameter Efficient Knowledge Distillation for Large Language Models
by: Yang, Runming, et al.
Published: (2024)

Rethinking Kullback-Leibler Divergence in Knowledge Distillation for Large Language Models
by: Wu, Taiqiang, et al.
Published: (2024)

Internalize the Temperature: On-Policy Self-Distillation as Policy Reheater for Reinforcement Learning
by: Yang, Xuewei, et al.
Published: (2026)

Hierarchical Memory for High-Efficiency Long-Term Reasoning in LLM Agents
by: Sun, Haoran, et al.
Published: (2025)

Distributional Clarity: The Hidden Driver of RL-Friendliness in Large Language Models
by: Sun, Shaoning, et al.
Published: (2026)

Think Parallax: Solving Multi-Hop Problems via Multi-View Knowledge-Graph-Based Retrieval-Augmented Generation
by: Liu, Jinliang, et al.
Published: (2025)

First Token Probability Guided RAG for Telecom Question Answering
by: Chen, Tingwei, et al.
Published: (2025)

S2J: Bridging the Gap Between Solving and Judging Ability in Generative Reward Models
by: Sun, Shaoning, et al.
Published: (2025)

Quagmires in SFT-RL Post-Training: When High SFT Scores Mislead and What to Use Instead
by: Kang, Feiyang, et al.
Published: (2025)

A Novel Self-Evolution Framework for Large Language Models
by: Sun, Haoran, et al.
Published: (2025)

Preference-Aware Memory Update for Long-Term LLM Agents
by: Sun, Haoran, et al.
Published: (2025)

An Empirical Study of SFT-DPO Interaction and Parameterization in Small Language Models
by: Feng, Yuming, et al.
Published: (2026)

Alignment-Enhanced Decoding:Defending via Token-Level Adaptive Refining of Probability Distributions
by: Liu, Quan, et al.
Published: (2024)

A Survey on the Honesty of Large Language Models
by: Li, Siheng, et al.
Published: (2024)

Minor SFT loss for LLM fine-tune to increase performance and reduce model deviation
by: Xie, Shiming, et al.
Published: (2024)

Rho-1: Not All Tokens Are What You Need
by: Lin, Zhenghao, et al.
Published: (2024)

A Survey of Reasoning in Autonomous Driving Systems: Open Challenges and Emerging Paradigms
by: Yu, Kejin, et al.
Published: (2026)

TPA: Next Token Probability Attribution for Detecting Hallucinations in RAG
by: Lu, Pengqian, et al.
Published: (2025)

Selective Preference Optimization via Token-Level Reward Function Estimation
by: Yang, Kailai, et al.
Published: (2024)

Can We Trust LLMs on Memristors? Diving into Reasoning Ability under Non-Ideality
by: Wu, Taiqiang, et al.
Published: (2026)

Edge-free but Structure-aware: Prototype-Guided Knowledge Distillation from GNNs to MLPs
by: Wu, Taiqiang, et al.
Published: (2023)

MaskCD: Mitigating LVLM Hallucinations by Image Head Masked Contrastive Decoding
by: Deng, Jingyuan, et al.
Published: (2025)

Critical Tokens Matter: Token-Level Contrastive Estimation Enhances LLM's Reasoning Capability
by: Lin, Zicheng, et al.
Published: (2024)

The Art of Efficient Reasoning: Data, Reward, and Optimization
by: Wu, Taiqiang, et al.
Published: (2026)

CreativeBench: Benchmarking and Enhancing Machine Creativity via Self-Evolving Challenges
by: Wang, Zi-Han, et al.
Published: (2026)

Do Not Let Low-Probability Tokens Over-Dominate in RL for LLMs
by: Yang, Zhihe, et al.
Published: (2025)

Reward Modeling from Natural Language Human Feedback
by: Wang, Zongqi, et al.
Published: (2026)

AceReason-Nemotron 1.1: Advancing Math and Code Reasoning through SFT and RL Synergy
by: Liu, Zihan, et al.
Published: (2025)

An Expert is Worth One Token: Synergizing Multiple Expert LLMs as Generalist via Expert Token Routing
by: Chai, Ziwei, et al.
Published: (2024)

Towards Token-Level Text Anomaly Detection
by: Cao, Yang, et al.
Published: (2026)

Analyzing the Effects of Supervised Fine-Tuning on Model Knowledge from Token and Parameter Levels
by: Ye, Junjie, et al.
Published: (2025)

Efficient Switchable Safety Control in LLMs via Magic-Token-Guided Co-Training
by: Si, Jianfeng, et al.
Published: (2025)

CollectiveSFT: Scaling Large Language Models for Chinese Medical Benchmark with Collective Instructions in Healthcare
by: Zhu, Jingwei, et al.
Published: (2024)

PDC & DM-SFT: A Road for LLM SQL Bug-Fix Enhancing
by: Duan, Yiwen, et al.
Published: (2024)

CoWork-X: Experience-Optimized Co-Evolution for Multi-Agent Collaboration System
by: Lin, Zexin, et al.
Published: (2026)

Beyond SFT-to-RL: Pre-alignment via Black-Box On-Policy Distillation for Multimodal RL
by: Wang, Sudong, et al.
Published: (2026)