Understanding LLM Parameter Update Sparsity through the Lens of Fisher

arXiv:2609.36262 · cs.LG · Submitted 2026-09-28 · Read on arXiv

cs.LG

Submitted: 2026-09-28

Updated: 2026-09-28

Terminology

• Sparsity — 1× in this paper · explained in 4 episodes

Sources

• The Multiple Ticket Hypothesis: Random Sparse Subnetworks Suffice for RLVR
• DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning
• Universal Neurons in GPT2 Language Models
• Let's Verify Step by Step
• Understanding R1-Zero-Like Training: A Critical Perspective
• RL Forgets! Towards Continual Policy Optimization
• Understanding and Exploiting Weight Update Sparsity for Communication-Efficient Distributed RL
• Do We Need Adam? Surprisingly Strong and Sparse Reinforcement Learning with SGD in LLMs
• Information-Geometric Forward Policy Training in GFlowNets
• DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models
• Qwen3 Technical Report
• Dense Supervision, Sparse Updates: On the Sparsity and Geometry of On-Policy Distillation
• The Path Not Taken: RLVR Provably Learns Off the Principals

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