The paper proposes Predictive Routing Replay (PR2) to stabilize reinforcement learning on Mixture of Experts (MoE) LLMs by predicting and incorporating short-horizon router evolution during training a…

The paper proposes BaSE, a multi-armed bandit approach, to optimally allocate a fixed budget of LLM calls across parallel evolutionary search trajectories, significantly improving mean fitness and rel…

The paper proposes using distributional Reinforcement Learning (RL) to stabilize learning in chaotic dynamical systems by optimizing the smooth evolution of the return distribution rather than individ…

The paper proposes S2L-PO, a framework that uses smaller, naturally diverse models as structured explorers to enhance the policy-level diversity and performance of larger language models during traini…

The paper introduces Straggler-Aware Group Control (SAGC), a dynamic group-size controller that optimizes synchronous on-policy RL training by adapting group size to minimize delays caused by slow rol…

This paper proposes two horizon-control strategies, Progressive OPD (POPD) and Truncated OPD (TOPD), demonstrating that full rollouts are often unnecessary for On-Policy Distillation, leading to signi…

The paper introduces a diagnostic-driven iterative refinement process for improving LLM-generated reward functions in sparse, structured reinforcement learning tasks, significantly boosting agent perf…

The paper introduces Prompted Policy Optimization (PromptPO), an LLM-based method that successfully optimizes policies for various sequential RL tasks, demonstrating that LLMs can replace classical RL…

This paper proposes a new imitation learning algorithm called DistIL that uses distributional feedback to improve policy improvement and regret guarantees.

DriftQL introduces a novel, efficient offline RL method that combines a drift-based behavioral regularizer with critic-driven policy improvement, achieving state-of-the-art performance while maintaini…

Lodestar is a novel online learning-based request routing system that significantly improves LLM inference efficiency by dynamically assigning incoming requests to the optimal GPU instance to minimize…

ESPO is a novel reinforcement learning algorithm that detects trajectory failure in large language models and terminates rollouts early, significantly improving performance on mathematical reasoning b…

EchoRL proposes a lightweight module to exploit valuable learning signals from advantage-degenerated rollouts in Reinforcement Learning with Verifiable Rewards (RLVR), significantly improving LLM post…

ZipRL introduces an adaptive context compression framework that significantly improves the performance and efficiency of LLMs in complex, multi-turn agent tasks by combining multi-granularity compress…

The paper introduces the Terminal Representation (TR), a novel, lower-dimensional, and structurally distinct formulation for encoding reward-weighted trajectories in RL that bypasses the need for eige…

ADWIN introduces an adaptive window framework for on-policy distillation (OPD) that efficiently manages the supervision horizon by training on short, teacher-anchored prefixes while using delayed full…

OrcaRouter is a production-ready LLM router that uses a hybrid offline-online learning approach to efficiently select the best large language model for an incoming query, achieving high accuracy at lo…

The Parameterized Diffusion Policy (PDP) framework transforms diffusion models from general stochastic generators into precise, steerable tools for learning and adapting complex robotic behaviors by e…

The paper addresses the failure of fixed-price inference in resource-constrained pricing controllers by developing a target-aware controller that tracks local densities and provides certified, shrinki…

This paper demonstrates that a complex deep reinforcement learning policy for power grid control can be successfully distilled into a lightweight, auditable decision tree and random forest surrogate t…