Image de couverture de livre

SFT Doesn't Always Hurt General Capabilities: Revisiting Domain-Specific Fine-Tuning in LLMs

Fuente: arXiv

Enregistré dans:

Détails bibliographiques
Auteurs principaux:	Lin, Jiacheng, Wang, Zhongruo, Qian, Kun, Wang, Tian, Srinivasan, Arvind, Zeng, Hansi, Jiao, Ruochen, Zhou, Xie, Gesi, Jiri, Wang, Dakuo, Guo, Yufan, Zhong, Kai, Zhang, Weiqi, Sanghavi, Sujay, Chen, Changyou, Yun, Hyokun, Li, Lihong
Format:	Preprint
Publié:	2025
Sujets:	Computation and Language
Accès en ligne:	Acceder al recurso
Tags:	Ajouter un tag Pas de tags, Soyez le premier à ajouter un tag!

Documents similaires

Language Complexity and Speech Recognition Accuracy: Orthographic Complexity Hurts, Phonological Complexity Doesn't
par: Taguchi, Chihiro, et autres
Publié: (2024)

Bayesian Elicitation with LLMs: Model Size Helps, Extra "Reasoning" Doesn't Always
par: Hobor, Luka, et autres
Publié: (2026)

Library Designs Revisited: What Works--What Doesn't.
par: Metz, T. John, et autres
Publié: (1987)

Consciousness Doesn't Do That
par: Matthias Michel
Publié: (2026)

The Apple Doesn't Fall Far From the Tree: A Three‐Level Examination of the Corporate Hypocrisy Trickle‐Down Effect
par: Mingchuan Yu, et autres
Publié: (2026)

Chapter When It Doesn't Go to Plan
par: Kim, Seoyeon | https://orcid.org/0000-0002-2018-9520, et autres
Publié: (2026)

Why Johnny Can Read...But Doesn't
par: Landy, Sarah
Publié: (1977)

Teach AI What It Doesn't Know
par: Sean Du
Publié: (2026)

When the Conversation Doesn't Go Your Way
Publié: (2024)

Sometimes the Roof Doesn't Just Leak, It Caves In!
par: Curran, Charles, et autres
Publié: (1999)

Revisiting Intermediate-Layer Matching in Knowledge Distillation: Layer-Selection Strategy Doesn't Matter (Much)
par: Yu, Zony, et autres
Publié: (2025)

Doesn't Everyone Have Rights to a Learner's Permit?
par: Gehrig, Jody, et autres
Publié: (2009)

A Teacher's Influence Doesn't Apply Only to Students
par: Coatney, Sharon
Publié: (2005)

Staff Development in Libraries: Why It Frequently Doesn't Take.
par: Shaughnessy, Thomas W.
Publié: (1988)

REPA Works Until It Doesn't: Early-Stopped, Holistic Alignment Supercharges Diffusion Training
par: Wang, Ziqiao, et autres
Publié: (2025)

Context-Free Synthetic Data Mitigates Forgetting
par: Bansal, Parikshit, et autres
Publié: (2025)

Enabling Approximate Joint Sampling in Diffusion LMs
par: Bansal, Parikshit, et autres
Publié: (2025)

Beyond Self-learned Attention: Mitigating Attention Bias in Transformer-based Models Using Attention Guidance
par: Gesi, Jiri, et autres
Publié: (2024)

The Clock Doesn't Close: Euler-Mascheroni as Torus Non-Closure
par: David Jan Lowder, et autres
Publié: (2026)

Against Parfitian Aggregation: Why Reductionism Doesn't Save Utilitarianism
par: Biagi, Tommaso
Publié: (2026)

Semantics at an Angle: When Cosine Similarity Works Until It Doesn't
par: You, Kisung
Publié: (2025)

Library Learning Doesn't: The Curious Case of the Single-Use "Library"
par: Berlot-Attwell, Ian, et autres
Publié: (2024)

Korea's Solidarity with the Global South (to Which It Didn't and Doesn't Belong)*
par: Suweon Kim
Publié: (2025)

Something There Is That Doesn't Love a Computer (Nor Hate It Either).
par: Friedman, Fred T.
Publié: (1984)

Shop-R1: Rewarding LLMs to Simulate Human Behavior in Online Shopping via Reinforcement Learning
par: Zhang, Yimeng, et autres
Publié: (2025)

One LR Doesn't Fit All: Heavy-Tail Guided Layerwise Learning Rates for LLMs
par: He, Di, et autres
Publié: (2026)

Bell Doesn't Play Dice! The Classical Mechanical Origin of Quantum Entanglement
par: Dolce, Donatello
Publié: (2025)

Recurrent Off-Policy Deep Reinforcement Learning Doesn't Have to be Slow
par: Clark, Tyler, et autres
Publié: (2025)

ChatGPT Doesn't Trust Chargers Fans: Guardrail Sensitivity in Context
par: Li, Victoria R., et autres
Publié: (2024)

Linux Kernel Recency Matters, CVE Severity Doesn't, and History Fades
par: Przymus, Piotr, et autres
Publié: (2026)

Testing Autonomous Driving Systems -- What Really Matters and What Doesn't
par: Li, Changwen, et autres
Publié: (2025)

Explorations of the Softmax Space: Knowing When the Neural Network Doesn't Know
par: Sikar, Daniel, et autres
Publié: (2025)

Reinforcement Learning for Reasoning in Small LLMs: What Works and What Doesn't
par: Dang, Quy-Anh, et autres
Publié: (2025)

When More Data Doesn't Help: Limits of Adaptation in Multitask Learning
par: Hanneke, Steve, et autres
Publié: (2026)

It Doesn't Matter How Many "Doses": One-Shots Aren't Cures
par: Santamaria, Michele, et autres
Publié: (2022)

The Usual Doesn't Work: Why We Need Problem-Based Learning
par: Spence, Larry
Publié: (2004)

In-Service and the School Library Media Specialist: What Works and What Doesn't.
par: Turner, Philip M.
Publié: (1988)

"Something Comes through or It Doesn't": Intensive Reading in Post-Qualitative Inquiry
par: Maggie MacLure
Publié: (2024)

Ice Cream Doesn't Cause Drowning: Benchmarking LLMs Against Statistical Pitfalls in Causal Inference
par: Du, Jin, et autres
Publié: (2025)

Multi-Agent-as-Judge: Aligning LLM-Agent-Based Automated Evaluation with Multi-Dimensional Human Evaluation
par: Chen, Jiaju, et autres
Publié: (2025)