R-Diverse: Mitigating Diversity Illusion in Self-Play LLM Training

• URL: http://arxiv.org/abs/2602.13103v2
• Date: Mon, 16 Feb 2026 09:38:20 GMT
• Title: R-Diverse: Mitigating Diversity Illusion in Self-Play LLM Training
• Abstract summary: Self-play LLM reasoning through an iterative Challenger bootstrapr loop.<n>R-Diverse sustains gains over more iterations and consistently outperforms prior self-play methods.
• Score: 65.13759782915164
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Self-play bootstraps LLM reasoning through an iterative Challenger-Solver loop: the Challenger is trained to generate questions that target the Solver's capabilities, and the Solver is optimized on the generated data to expand its reasoning skills. However, existing frameworks like R-Zero often exhibit non-sustained improvement, where early gains degrade as self-play continues. We identify a key failure mode, Diversity Illusion, where the Solver's training signals appear diverse yet collapse into recurring underlying patterns. It manifests as (1) Local Diversity Illusion, where diversity is enforced only within-batch, inducing cross-iteration mode cycling; and (2) Surface Diversity Illusion, where questions vary superficially but require near-identical reasoning skills. To mitigate them, we propose R-Diverse with two aligned innovations: Memory-Augmented Penalty (MAP), which uses a persistent memory bank to discourage recycling across iterations, and Skill-Aware Measurement (SAM), which evaluates diversity by the reasoning skills exercised rather than surface variation of questions. Across 10 math and general reasoning benchmarks, R-Diverse sustains gains over more iterations and consistently outperforms prior self-play methods. Code is available at https://github.com/Gengsheng-Li/R-Diverse.

Related papers

• ARISE-RL: Agentic Rubric-Grounded Iterative Self-Evolution with Reinforcement Learning [74.19730275669227]
  ARISE-RL is a novel full-cycle self-evolution framework for training open-ended agents.<n>We present Reward-Gated Self-Evolution Distillation (RG-SED), which distills a memory-augmented variant of the same policy back into itself.<n>We also present ECR-Bench, an expert-calibrated rubric benchmark suite covering single-tool deep research and multi-tool travel planning.
  arXiv  Detail & Related papers  (2026-09-01T10:54:13Z)
• Beyond Uncertainty: Multi-Solver Disagreement Rewards for Self-Evolving Reasoning Curricula [0.0]
  Self-evolving reasoning frameworks train a Challenger to generate questions exposing a solver's weaknesses.<n>Existing approaches use a single solver's sampling uncertainty as the Challenger's reward.<n>We propose a multi-solver disagreement reward using a heterogeneous ensemble varying in model capacity and sampling temperature.
  arXiv  Detail & Related papers  (2026-08-30T20:49:16Z)
• Thinking Before Matching: A Reinforcement Reasoning Paradigm Towards General Person Re-Identification [77.51622646061314]
  ReID-R is proposed as a novel reasoning-driven paradigm that achieves explicit identity understanding and reasoning.<n>Experiments on multiple ReID benchmarks demonstrate that ReID-R achieves competitive identity discrimination as superior methods.
  arXiv  Detail & Related papers  (2026-04-21T08:24:07Z)
• ReCALL: Recalibrating Capability Degradation for MLLM-based Composed Image Retrieval [64.14282916266998]
  Composed Image Retrieval aims to retrieve target images based on a hybrid query comprising a reference image and a modification text.<n>We propose ReCALL, a model-agnostic framework that follows a diagnose-generate-refine pipeline.<n>Experiments on CIRR and FashionIQ show that ReCALL consistently recalibrates degraded capabilities and achieves state-of-the-art performance.
  arXiv  Detail & Related papers  (2026-02-02T04:52:54Z)
• Rewarding the Rare: Uniqueness-Aware RL for Creative Problem Solving in LLMs [126.45104018441698]
  Reinforcement learning (RL) has become a central paradigm for post-training large language models (LLMs)<n>We argue that this failure stems from regularizing local token behavior rather than diversity over sets of solutions.<n>We propose Uniqueness-Aware Reinforcement Learning, a rollout-level objective that explicitly rewards correct solutions that exhibit rare high-level strategies.
  arXiv  Detail & Related papers  (2026-01-13T17:48:43Z)
• Outcome-based Exploration for LLM Reasoning [18.33816564983908]
  Reinforcement learning (RL) has emerged as a powerful method for improving the reasoning abilities of large language models.<n>We show that RL can reduce effective diversity even on the training set relative to the base model.<n>We propose outcome-based exploration, which assigns exploration bonuses according to final outcomes.
  arXiv  Detail & Related papers  (2025-09-08T17:52:56Z)
• R-Zero: Self-Evolving Reasoning LLM from Zero Data [47.8125954446991]
  Self-evolving Large Language Models (LLMs) offer a scalable path toward super-intelligence by autonomously generating, refining, and learning from their own experiences.<n>Existing methods for training such models still rely heavily on vast human-curated tasks and labels.<n>We introduce R-Zero, a fully autonomous framework that generates its own training data from scratch.
  arXiv  Detail & Related papers  (2025-08-07T03:38:16Z)
• RAG-R1 : Incentivize the Search and Reasoning Capabilities of LLMs through Multi-query Parallelism [10.288667305064065]
  Large Language Models (LLMs) have demonstrated remarkable capabilities across various tasks.<n>LLMs remain prone to generating hallucinated or outdated responses due to their static internal knowledge.<n>Recent advancements in Retrieval-Augmented Generation (RAG) methods have aimed to enhance models' search and reasoning capabilities.
  arXiv  Detail & Related papers  (2025-06-30T09:02:45Z)
• Recursive Introspection: Teaching Language Model Agents How to Self-Improve [30.086494067593268]
  We develop RISE: Recursive IntroSpEction, an approach for fine-tuning large language models. Our experiments show that RISE enables Llama2, Llama3, and Mistral models to improve themselves with more turns on math reasoning tasks.
  arXiv  Detail & Related papers  (2024-07-25T17:35:59Z)
• Flow of Reasoning: Training LLMs for Divergent Reasoning with Minimal Examples [12.48027669682156]
  Flow of Reasoning (FoR) aims at improving diversity with minimal data.<n>FoR formulates multi-step LLM reasoning as a Markovian flow on a DAG-structured reasoning graph.<n>Experiments show that, with limited training examples, FoR enables the discovery of diverse, creative, high-quality solutions.
  arXiv  Detail & Related papers  (2024-06-09T07:06:58Z)
• ReST meets ReAct: Self-Improvement for Multi-Step Reasoning LLM Agent [50.508669199496474]
  We develop a ReAct-style LLM agent with the ability to reason and act upon external knowledge. We refine the agent through a ReST-like method that iteratively trains on previous trajectories. Starting from a prompted large model and after just two iterations of the algorithm, we can produce a fine-tuned small model.
  arXiv  Detail & Related papers  (2023-12-15T18:20:15Z)
• Diversify Question Generation with Retrieval-Augmented Style Transfer [68.00794669873196]
  We propose RAST, a framework for Retrieval-Augmented Style Transfer. The objective is to utilize the style of diverse templates for question generation. We develop a novel Reinforcement Learning (RL) based approach that maximizes a weighted combination of diversity reward and consistency reward.
  arXiv  Detail & Related papers  (2023-10-23T02:27:31Z)
• Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection [74.51523859064802]
  We introduce a new framework called Self-Reflective Retrieval-Augmented Generation (Self-RAG) Self-RAG enhances an LM's quality and factuality through retrieval and self-reflection. It significantly outperforms state-of-the-art LLMs and retrieval-augmented models on a diverse set of tasks.
  arXiv  Detail & Related papers  (2023-10-17T18:18:32Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.