Publications & Preprints
From worst-case optimization and reinforcement learning to a combinatorial probability puzzle. An asterisk marks equal contribution.
Conference Papers
-
2026
Adaptive Scheduling Pipeline for Multi-Instance Asynchronous Reinforcement Learning
NeurIPS 2026, main track ACCEPTED
ASP is a bounded-lag asynchronous architecture for RL post-training of large language models. A queue-aware scheduling policy jointly reduces padding, skewness and sorting bubbles, enabling efficient multi-instance rollout. ASP improves end-to-end throughput by up to 3.22× and reduces iteration time by up to 76%, with no visible convergence degradation.
Preprints & Submitted
-
2026
Taylor Representations for Model-Free RL in Networked MDPs
Submitted to ICLR 2027
In networked MDPs, local-critic Taylor coefficients decay exponentially with the graph distance to the farthest agent involved, so discarding distant-agent coefficients yields scalable local Taylor representations. Building on them, we propose a model-free actor–critic algorithm with finite-sample guarantees for a linear LSTD critic. Across three control benchmarks, it matches or outperforms spectral baselines while scaling efficiently to large graphs.
-
2026
Last-Iterate Performance of Gradient Descent and Relaxed Proximal Point via s-Composability
Preprint
We show that the joint potential of s-composability, introduced to compose optimized stepsize schedules, also certifies sharp last-iterate performance. The constant schedule is s-composable at every horizon, which identifies the unique constant stepsize minimizing the worst-case final gradient norm of smooth convex gradient descent, and of relaxed proximal point through the Moreau envelope. We also obtain the exact last-iterate constant of the silver schedule.
-
2025
An Analytical Framework for the Levine Hats Problem: New Strategies, Bounds and Generalizations
Preprint
In the Levine hat problem, n ≥ 2 players must each identify a black hat on their own infinite stack, seeing only their teammates' stacks. Representing strategies as measurable functions, we construct a recursive strategy 𝒮5 that attains the conjectured optimum 7/20 for two players, and we show that with uncountably infinite stacks the optimal winning probability is exactly 1/2.
* Equal contribution.