yongliang-wu/DFT

Health v2 · maintenance only

Flags: no_releases no_license

Adoption not part of the score

1100 stars · 26 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

DFT (Dynamic Fine-Tuning) is the official implementation of an ICLR 2026 paper that improves Supervised Fine-Tuning of LLMs by dynamically rescaling the loss with token probabilities, rectifying the implicit reward structure of SFT gradients. It requires only a single-line change to standard SFT and improves generalization on math reasoning, code generation, and multi-modal benchmarks.

Use cases

• fine-tune an LLM with better generalization than standard SFT
• improve supervised fine-tuning for math reasoning tasks
• apply a one-line loss modification to an existing SFT pipeline
• use DFT as a streamlined alternative to offline RL fine-tuning
• reproduce ICLR 2026 paper results on models up to 7B parameters

When to choose

• you are fine-tuning LLMs on math or code datasets and want better generalization than vanilla SFT
• you want an RL-like generalization boost without running full RL training
• you are doing research on SFT vs RL training dynamics

When to avoid

• your fine-tuning domain is literary, financial, or otherwise outside math/code tasks where DFT is unproven
• you need a production-grade, licensed, well-supported training framework
• you work with models far beyond 7B parameters, which the paper did not evaluate

Facets

library · maturity active

llm-training machine-learning deep-learning large-language-models machine-learning deep-learning python sft fine-tuning reinforcement-learning research-paper iclr-2026 reward-rectification math-reasoning gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="yongliang-wu/DFT")