Pith. sign in

REVIEW 1 cited by

Return Capping: Sample-Efficient CVaR Policy Gradient Optimisation

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2504.20887 v2 pith:EZCTUFPZ submitted 2025-04-29 cs.LG cs.AI

Return Capping: Sample-Efficient CVaR Policy Gradient Optimisation

classification cs.LG cs.AI
keywords cvarproblemcappingdiscardingoptimisationpolicyreformulationresults
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
Share X Bluesky LinkedIn Reddit HN
read the original abstract

When optimising for conditional value at risk (CVaR) using policy gradients (PG), current methods rely on discarding a large proportion of trajectories, resulting in poor sample efficiency. We propose a reformulation of the CVaR optimisation problem by capping the total return of trajectories used in training, rather than simply discarding them, and show that this is equivalent to the original problem if the cap is set appropriately. We show, with empirical results in an number of environments, that this reformulation of the problem results in consistently improved performance compared to baselines. We have made all our code available here: https://github.com/HarryMJMead/cvar-return-capping.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Reward Redistribution for CVaR MDPs using a Bellman Operator on L-infinity

    cs.LG 2026-02 conditional novelty 5.0

    A shifted-value transformation turns static CVaR MDPs into a bounded, contracting Bellman operator with dense rewards, enabling discretized value iteration and Q-learning with explicit error bounds.