Learning Task Mixtures from Task Affinities: A Probabilistic Graphical Model for Supervised Fine-Tuning

arXiv:2507.12612 · cs.LG, cs.AI · Submitted 2025-07-16 · Read on arXiv

cs.LG, cs.AI

Submitted: 2025-07-16

Updated: 2026-09-28

Terminology

• Fine-Tuning — 2× in this paper · explained in 5 episodes
• Mutual Information — 1× in this paper · explained in 2 episodes
• Redundancy — 1× in this paper · explained in 2 episodes

Sources

• Improving Influence-based Instruction Tuning Data Selection for Balanced Learning of Diverse Capabilities
• Tulu 3: Pushing Frontiers in Open Language Model Post-Training
• Data Mixing Optimization for Supervised Fine-Tuning of Large Language Models
• RegMix: Data Mixture as Regression for Language Model Pre-training
• D2 Pruning: Message Passing for Balancing Diversity and Difficulty in Data Pruning
• Task Arithmetic in the Tangent Space: Improved Editing of Pre-Trained Models
• Multitask Prompted Training Enables Zero-Shot Task Generalization
• LLaMA: Open and Efficient Foundation Language Models
• Qwen2 Technical Report

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