paper-with-me

Papers

IRDS: Interpretable RLVR Data Selection via Verifier-Coupled Sparse Autoencoder Coverage

2026-05-27 · Yuhan Li, Mingxu Zhang, Dazhong Shen, Ying Sun arxiv

Reinforcement learning with verifiable rewards (RLVR) has become a key technique for en- hancing LLM reasoning, yet its data ineffi- ciency remains a major bottleneck. Existing methods address this problem only partially, each missing at least one of subset-level cov- erage, verifier signal use, or interpretability. To address this gap, we present IRDS (Inter- pretable RLVR Data Selection), which selects RLVR training instances on a sparse autoen- coder (SAE) cluster basis so the selection itself is auditable on recognizable problem motifs. To select instances the model both fails on and can still learn from, we introduce a verifier- coupled coverage objective on the SAE basis and solve it by greedy log-determinant max- imization. Experiments on three instruction- tuned models and six math reasoning bench- marks show that IRDS achieves the highest overall accuracy, exceeding the strongest base- line by +3.9/+4.0 pp on the two Qwen models and by +0.5 pp on Llama-3.1-8B, while run- ning an order of magnitude cheaper than the trajectory-based baseline.

📄 PDF Abstract BibTeX arXiv:2605.28247

Code (0)

등록된 구현이 없습니다.

Tasks

Reinforcement Learning

Similar Papers 제목 키워드 기반

Aletheia: What Makes RLVR For Code Verifiers Tick?

2026-01-17 · Vatsal Venkatkrishna, Indraneil Paul, Iryna Gurevych arxiv

Multi-domain thinking verifiers trained via Reinforcement Learning with Verifiable Rewards (RLVR) are a cornerstone of modern post-training. However, their adoption in code generation has lagged behind that of execution …

Reinforcement LearningCode Generation

PRIME: A Process-Outcome Alignment Benchmark for Verifiable Reasoning in Mathematics and Engineering

2026-02-12 · Xiangfeng Wang, Hangyu Guo, Yanlin Lai, Mitt Huang 외 arxiv

While model-based verifiers are essential for scaling Reinforcement Learning with Verifiable Rewards (RLVR), current outcome-centric verification paradigms primarily focus on the consistency between the final result and …

Reinforcement Learning

LLMs Gaming Verifiers: RLVR can Lead to Reward Hacking

2026-04-16 · Lukas Helff, Quentin Delfosse, David Steinmann, Ruben Härle 외 arxiv

As reinforcement Learning with Verifiable Rewards (RLVR) has become the dominant paradigm for scaling reasoning capabilities in LLMs, a new failure mode emerges: LLMs gaming verifiers. We study this phenomenon on inducti…

Reinforcement Learning

Before the Model Learns the Bug:Fuzzing RLVR Verifiers

2026-05-31 · Jaideep Ray arxiv

Reinforcement learning with verifiable rewards (RLVR) replaces human preference labels with executable reward functions such as math answer checkers, JSON tool-call validators, and code unit-test harnesses. That makes th…

Reinforcement Learning

Reinforcement Learning with Verifiable yet Noisy Rewards under Imperfect Verifiers

2025-10-01 · Xin-Qiang Cai, Wei Wang, Feng Liu, Tongliang Liu 외 arxiv

Reinforcement Learning with Verifiable Rewards (RLVR) replaces costly human labeling with automated verifiers. To reduce verifier hacking, many RLVR systems binarize rewards to $\{0,1\}$, but imperfect verifiers inevitab…

Reinforcement Learning