Pith. sign in

REVIEW 13 cited by

Tapered Off-Policy REINFORCE: Stable and efficient reinforcement learning for LLMs

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 2503.14286 v2 pith:HA273CXN submitted 2025-03-18 cs.LG

Tapered Off-Policy REINFORCE: Stable and efficient reinforcement learning for LLMs

classification cs.LG
keywords exampleslearningnegativeoff-policyperformancereinforcetaperedtraining
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
read the original abstract

We propose a new algorithm for fine-tuning large language models using reinforcement learning. Tapered Off-Policy REINFORCE (TOPR) uses an asymmetric, tapered variant of importance sampling to speed up learning while maintaining stable learning dynamics, even without the use of KL regularization. TOPR can be applied in a fully offline fashion, allows the handling of positive and negative examples in a unified framework, and benefits from the implementational simplicity that is typical of Monte Carlo algorithms. We demonstrate the effectiveness of our approach with a series of experiments on the GSM8K and MATH reasoning benchmarks, finding performance gains for training both a model for solution generation and as a generative verifier. We show that properly leveraging positive and negative examples alike in the off-policy regime simultaneously increases test-time accuracy and training data efficiency, all the while avoiding the ``wasted inference'' that comes with discarding negative examples. We find that this advantage persists over multiple iterations of training and can be amplified by dataset curation techniques, enabling us to match 70B-parameter model performance with 8B language models. As a corollary to this work, we find that REINFORCE's baseline parameter plays an important and unexpected role in defining dataset composition in the presence of negative examples, and is consequently critical in driving off-policy performance.

discussion (0)

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

Forward citations

Cited by 13 Pith papers

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

  1. Rethinking Muon Beyond Pretraining: Spectral Failures and High-Pass Remedies for VLA and RLVR

    cs.LG 2026-05 conditional novelty 7.0

    Pion modifies Muon's Newton-Schulz iterations into a controllable high-pass filter that anchors dominant singular values at 1 while suppressing noisy tails, outperforming Muon and AdamW in VLA and RLVR regimes.

  2. Freshness-Aware Prioritized Experience Replay for LLM/VLM Reinforcement Learning

    cs.CL 2026-04 unverdicted novelty 7.0

    Freshness-Aware PER augments prioritized experience replay with exponential age decay based on effective sample size to enable successful reuse of trajectories in LLM and VLM reinforcement learning, outperforming on-p...

  3. UP: Unbounded Positive Asymmetric Optimization for Breaking the Exploration-Stability Dilemma

    cs.LG 2026-07 conditional novelty 6.0

    Replacing the importance sampling ratio with a stop-gradient self-anchored ratio for positive advantages yields unclipped, REINFORCE-equivalent gradients that improve exploration without training instability.

  4. What are Key Factors for Updates in RL for LLM Reasoning?

    cs.CL 2026-06 unverdicted novelty 6.0

    Theoretical analysis of RLVR update dynamics leads to ACPO, an adaptive clipping method that outperforms DAPO and CISPO on reasoning benchmarks with 3B and 7B models.

  5. Rollout-Level Advantage-Prioritized Experience Replay for GRPO

    cs.LG 2026-06 conditional novelty 6.0

    Rollout-level advantage-prioritized experience replay for GRPO recycles high-advantage individual rollouts with age eviction and fresh-anchored batches to outperform standard GRPO on math benchmarks, with gains increa...

  6. Extreme Region Policy Distillation

    cs.LG 2026-05 unverdicted novelty 6.0

    ERPD decouples aggressive off-policy optimization on fixed trajectories from trust-region distillation to achieve comparable or better LLM performance with substantially smaller KL divergence.

  7. How Off-Policy Can GRPO Be? Mu-GRPO for Efficient LLM Reinforcement Learning

    cs.LG 2026-05 conditional novelty 6.0

    Mu-GRPO enables substantially more off-policy GRPO training for LLMs via relaxed clipping and negative-advantage veto in large staged batches, matching standard GRPO performance at ~2x training speed.

  8. Missing Old Logits in Asynchronous Agentic RL: Semantic Mismatch and Repair Methods for Off-Policy Correction

    cs.LG 2026-05 unverdicted novelty 6.0

    Missing old logits in async agentic RL entangle discrepancy and staleness terms in PPO off-policy correction; exact acquisition methods and revised PPO-EWMA restore decoupled updates with reported gains in speed and p...

  9. Modularized Reinforcement Learning on LLMs: From MDP Creation to Exploration and Learning

    cs.LG 2026-06 unverdicted novelty 5.0

    Survey mapping RL techniques onto LLM training and highlighting gaps in value-based, off-policy, and bootstrapping methods.

  10. Trust Region On-Policy Distillation

    cs.LG 2026-05 unverdicted novelty 5.0

    TrOPD stabilizes on-policy distillation for LLMs with trust-region learning, outlier estimation, and off-policy guidance, outperforming prior OPD methods on reasoning and code benchmarks.

  11. Ratio-Variance Regularized Policy Optimization

    cs.LG 2026-05 unverdicted novelty 5.0

    R²VPO uses ratio-variance regularization as a distributional soft brake on policy updates, claiming better performance than PPO on math reasoning and robotic control without hard clipping.

  12. Polychromic Objectives for Reinforcement Learning

    cs.LG 2025-09 unverdicted novelty 5.0

    Introduces polychromic objectives adapted into PPO via vine sampling and modified advantages, showing higher success rates and better coverage under perturbations on BabyAI, Minigrid, and algorithmic tasks.

  13. Reinforcement Learning from Human Feedback

    cs.LG 2025-04 unverdicted novelty 2.0

    The book introduces the origins, mathematical setup, and optimization stages of RLHF including reward modeling, reinforcement learning, rejection sampling, and direct alignment algorithms.