Beyond Full-Model Rollback: AuroSFT for Adapter-State Multi-Task Fine-Tuning
Yue Han, Ziniu Liu
cs.LG
Submitted: 2026-08-05
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge
- Training Verifiers to Solve Math Word Problems
- QLoRA: Efficient Finetuning of Quantized LLMs
- AuroRA: Breaking Low-Rank Bottleneck of LoRA with Nonlinear Mapping
- The Llama 3 Herd of Models
- LoRA: Low-Rank Adaptation of Large Language Models
- mSFT: Addressing Dataset Mixtures Overfitting Heterogeneously in Multi-task SFT
- Prefix-Tuning: Optimizing Continuous Prompts for Generation
- Data Mixing Optimization for Supervised Fine-Tuning of Large Language Models
- DoRA: Weight-Decomposed Low-Rank Adaptation
- Qwen2.5 Technical Report
- DaMo: Data Mixing Optimizer in Fine-tuning Multimodal LLMs for Mobile Phone Agents
- DynamixSFT: Dynamic Mixture Optimization of Instruction Tuning Collections
- LoRA-GA: Low-Rank Adaptation with Gradient Approximation
- SFTMix: Elevating Language Model Instruction Tuning with Mixup Recipe
- AdaLoRA: Adaptive Budget Allocation for Parameter-Efficient Fine-Tuning
- Instruction Tuning for Large Language Models: A Survey
Related papers
- Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
- AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
- Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
- Convergence issues in Relational Concept Analysis based on AOC-posets
- Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
- Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks