FSPO: Policy-Consistent Risk and Pareto-Feasible Control for Budgeted LLM RL Post-Training

cs.AI, cs.CL, cs.LG

Submitted: 2026-10-02

Updated: 2026-10-02

Comments: 40 pages, 4 figures

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Sources

• A Gentle Introduction to Conformal Prediction and Distribution-Free Uncertainty Quantification
• FrugalGPT: How to Use Large Language Models While Reducing Cost and Improving Performance
• Cutting AI Datacenter Energy with Reinforcement Learning: Measured Power Control of LLM Training from One GPU to the Fleet
• D4RL: Datasets for Deep Data-Driven Reinforcement Learning
• Population Based Training of Neural Networks
• RewardBench: Evaluating Reward Models for Language Modeling
• Proximal Policy Optimization Algorithms
• Learning to be Safe: Deep RL with a Safety Critic
• Where Should RL Post-Training Compute Go? Model Size, Search, Learning, and Feedback
• Using Common Random Numbers for Simulation-based Planning with Rollouts
• DAPO: An Open-Source LLM Reinforcement Learning System at Scale
• ProcessBench: Identifying Process Errors in Mathematical Reasoning
• Fine-Tuning Language Models from Human Preferences

Related papers

• MAVEN-T: Reinforced Heterogeneous Distillation for Real-Time Multi-Agent Trajectory Prediction
• Model Discovery Agent: LLM-assisted Bayesian experiment design for data-efficient discovery of mechanistic world models
• Agents' Last Exam
• The Clinician's Veto: Navigating Trust, Liability, and Uncertainty in Autonomous AI Prescribing
• MindHelper: Closed-Loop Embodied Mental-State Reasoning for Precision Intervention
• Incumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation Systems