Copertina

Reconciling Contradictory Views on the Effectiveness of SFT in LLMs: An Interaction Perspective

Fuente: arXiv

Salvato in:

Dettagli Bibliografici
Autori principali:	Zhang, Junpeng, Cheng, Lei, Zhang, Guoxi, Cai, Hua, Xu, Qing, Zhang, Quanshi
Natura:	Preprint
Pubblicazione:	2026
Soggetti:	Artificial Intelligence
Accesso online:	Acceder al recurso
Tags:	Aggiungi Tag Nessun Tag, puoi essere il primo ad aggiungerne!!

Documenti analoghi

Revisiting Generalization Power of a DNN in Terms of Symbolic Interactions
di: Cheng, Lei, et al.
Pubblicazione: (2025)

Randomness of Low-Layer Parameters Determines Confusing Samples in Terms of Interaction Representations of a DNN
di: Zhang, Junpeng, et al.
Pubblicazione: (2025)

Two-Phase Dynamics of Interactions Explains the Starting Point of a DNN Learning Over-Fitted Features
di: Zhang, Junpeng, et al.
Pubblicazione: (2024)

Technical Report: Quantifying and Analyzing the Generalization Power of a DNN
di: He, Yuxuan, et al.
Pubblicazione: (2025)

Towards the Dynamics of a DNN Learning Symbolic Interactions
di: Ren, Qihan, et al.
Pubblicazione: (2024)

Rethinking Generalization in Reasoning SFT: A Conditional Analysis on Optimization, Data, and Model Capability
di: Ren, Qihan, et al.
Pubblicazione: (2026)

Technical Note: Defining and Quantifying AND-OR Interactions for Faithful and Concise Explanation of DNNs
di: Li, Mingjie, et al.
Pubblicazione: (2023)

Attribution Explanations for Deep Neural Networks: A Theoretical Perspective
di: Deng, Huiqi, et al.
Pubblicazione: (2025)

Does a Neural Network Really Encode Symbolic Concepts?
di: Li, Mingjie, et al.
Pubblicazione: (2023)

End-to-End Neuro-Symbolic Reinforcement Learning with Textual Explanations
di: Luo, Lirui, et al.
Pubblicazione: (2024)

Layerwise Change of Knowledge in Neural Networks
di: Cheng, Xu, et al.
Pubblicazione: (2024)

Dissecting Dissonance: Benchmarking Large Multimodal Models Against Self-Contradictory Instructions
di: Gao, Jin, et al.
Pubblicazione: (2024)

Debunk the Myth of SFT Generalization
di: Lin, Xiaofeng, et al.
Pubblicazione: (2025)

Disentangling Regional Primitives for Image Generation
di: Chen, Zhengting, et al.
Pubblicazione: (2024)

FedTreeLoRA: Reconciling Statistical and Functional Heterogeneity in Federated LoRA Fine-Tuning
di: Bian, Jieming, et al.
Pubblicazione: (2026)

Implicit Reward as the Bridge: A Unified View of SFT and DPO Connections
di: Wang, Bo, et al.
Pubblicazione: (2025)

RL Squeezes, SFT Expands: A Comparative Study of Reasoning LLMs
di: Matsutani, Kohsei, et al.
Pubblicazione: (2025)

The Interaction Bottleneck of Deep Neural Networks: Discovery, Proof, and Modulation
di: Deng, Huiqi, et al.
Pubblicazione: (2025)

Quantifying In-Context Reasoning Effects and Memorization Effects in LLMs
di: Lou, Siyu, et al.
Pubblicazione: (2024)

Towards Attributions of Input Variables in a Coalition
di: Zheng, Xinhao, et al.
Pubblicazione: (2023)

Red Teaming Language Models for Processing Contradictory Dialogues
di: Wen, Xiaofei, et al.
Pubblicazione: (2024)

Can LLMs Reconcile Knowledge Conflicts in Counterfactual Reasoning
di: Yamin, Khurram, et al.
Pubblicazione: (2025)

A Unified and Stable Risk Minimization Framework for Weakly Supervised Learning with Theoretical Guarantees
di: Zhang, Miao, et al.
Pubblicazione: (2025)

Cost-Sensitive Unbiased Risk Estimation for Multi-Class Positive-Unlabeled Learning
di: Zhang, Miao, et al.
Pubblicazione: (2025)

VickreyFeedback: Cost-efficient Data Construction for Reinforcement Learning from Human Feedback
di: Zhang, Guoxi, et al.
Pubblicazione: (2024)

TEMPLE: Incentivizing Temporal Understanding of Video Large Language Models via Progressive Pre-SFT Alignment
di: Li, Shicheng, et al.
Pubblicazione: (2025)

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

Explaining Generalization Power of a DNN Using Interactive Concepts
di: Zhou, Huilin, et al.
Pubblicazione: (2023)

Utilizing Autoregressive Networks for Full Lifecycle Data Generation of Rolling Bearings for RUL Prediction
di: Wang, Junliang, et al.
Pubblicazione: (2024)

MVR: Multi-view Video Reward Shaping for Reinforcement Learning
di: Luo, Lirui, et al.
Pubblicazione: (2026)

Complementing reinforcement learning with SFT through logit averaging in the post training of LLMs
di: Gan, Xingwei, et al.
Pubblicazione: (2026)

mSFT: Addressing Dataset Mixtures Overfitting Heterogeneously in Multi-task SFT
di: Koh, Woosung, et al.
Pubblicazione: (2026)

How Much Backtracking is Enough? Exploring the Interplay of SFT and RL in Enhancing LLM Reasoning
di: Cai, Hongyi James, et al.
Pubblicazione: (2025)

Getting More Juice Out of the SFT Data: Reward Learning from Human Demonstration Improves SFT for LLM Alignment
di: Li, Jiaxiang, et al.
Pubblicazione: (2024)

Image Generation from Contextually-Contradictory Prompts
di: Huberman, Saar, et al.
Pubblicazione: (2025)

A Game-Theoretic Negotiation Framework for Cross-Cultural Consensus in LLMs
di: Zhang, Guoxi, et al.
Pubblicazione: (2025)

Defining and Extracting generalizable interaction primitives from DNNs
di: Chen, Lu, et al.
Pubblicazione: (2024)

Where We Have Arrived in Proving the Emergence of Sparse Symbolic Concepts in AI Models
di: Ren, Qihan, et al.
Pubblicazione: (2023)

Cost-Effective Communication: An Auction-based Method for Language Agent Interaction
di: Fan, Yijia, et al.
Pubblicazione: (2025)

Post-Training is About States, Not Tokens: A State Distribution View of SFT, RL, and On-Policy Distillation
di: Nie, Dong
Pubblicazione: (2026)