mlsanityData to 7 Oct 2026

GASP: Guided Asymmetric Self-Play For Coding LLMs

arXiv 2603.15957 ↗First posted 16 Mar 2026cs.LGcited by 4

On the map: Synthetic data and self-improvement for LLMs → Self-play RL for reasoning

Abstract (arXiv)

Asymmetric self-play has emerged as a promising paradigm for post-training large language models, where a teacher continually generates questions for a student to solve at the edge of the student's learnability. Although these methods promise open-ended data generation bootstrapped from no human data, they suffer from one major problem: not all problems that are hard to solve are interesting or informative to improve the overall capabilities of the model. Current asymmetric self-play methods are goal-agnostic with no real grounding. We propose Guided Asymmetric Self-Play (GASP), where grounding is provided by real-data goalpost questions that are identified to pose a hard exploration challenge to the model. During self-play, the teacher first generates an easier variant of a hard question, and then a harder variant of that easier question, with the goal of gradually closing the gap to the goalpost throughout training. Doing so, we improve pass@20 on LiveCodeBench (LCB) by 2.5% over unguided asymmetric self-play, and through the curriculum constructed by the teacher, we manage to solve hard goalpost questions that remain out of reach for all baselines.

Key facts. Cited by 4 papers in the map (data through 7 Oct 2026). Placed in Self-play RL for reasoning, one of 13 directions in Synthetic data and self-improvement for LLMs. Builds on 20 of its 23 references that are in the map. First posted on arXiv 16 Mar 2026.

Builds on

20 of 23 references.

Closest ancestors · 3

Same direction, newest first

  1. 2025Language Self-Play For Data-Free Training9 Sep
  2. 2025Absolute Zero: Reinforced Self-play Reasoning with Zero Data6 May
  3. 2025STP: Self-play LLM Theorem Provers with Iterative Conjecturing and Proving31 Jan

Other references · 17

  1. 2025Qwen3 Technical Reportcited by 8,195
  2. 2021Evaluating Large Language Models Trained on Codecited by 8,053
  3. 2023Efficient Memory Management for Large Language Model Serving with PagedAttentioncited by 5,767
  4. 2021Program Synthesis with Large Language Modelscited by 3,509
  5. 2024HybridFlow: A Flexible and Efficient RLHF Frameworkcited by 1,980
  6. 2024LiveCodeBench: Holistic and Contamination Free Evaluation of Large Language Models for Codecited by 1,906
  7. 2023Is Your Code Generated by ChatGPT Really Correct? Rigorous Evaluation of Large Language Models for Code Generationcited by 1,332
  8. 2022Evolving Curricula with Regret-Based Environment Designcited by 166
  9. 2017Intrinsic Motivation and Automatic Curricula via Asymmetric Self-Playcited by 139
  10. 2025SPICE: Self-Play In Corpus Environments Improves Reasoningcited by 78
  11. 2021Asymmetric self-play for automatic goal discovery in robotic manipulationcited by 71
  12. 2024Learning Formal Mathematics From Intrinsic Motivationcited by 45
5 more
  1. 2025Guided Self-Evolving LLMs with Minimal Human Supervisioncited by 33
  2. 2024No Regrets: Investigating and Improving Regret Approximations for Curriculum Discoverycited by 29
  3. 2024Scalable Reinforcement Post-Training Beyond Static Human Prompts: Evolving Alignment via Asymmetric Self-Playcited by 21
  4. 2026Teaching Models to Teach Themselves: Reasoning at the Edge of Learnabilitycited by 13
  5. 2025Better LLM Reasoning via Dual-Playcited by 7

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 4.

4 papers.

  1. 2026Scaling Self-Play with Self-Guidancecited by 9
  2. 2026SPADE: Self-Play in Adaptive Synthetic Executable Environmentscited by 6
  3. 2026FrontierSmith: Synthesizing Open-Ended Coding Problems at Scalecited by 3
  4. 2026CompassPlay: Rewarding the Proposer for Where It Moves the Solvercited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025Self-Questioning Language Models
  2. 2026Bootstrapping Post-training Signals for Open-ended Tasks via Rubric-based Self-play on Pre-training Text
  3. 2026DARC: Decoupled Asymmetric Reasoning Curriculum for LLM Evolution
  4. 2025Learning to Solve and Verify: A Self-Play Framework for Code and Test Generation
  5. 2026SCOPE: Self-Play via Co-Evolving Policies for Open-Ended Tasks
  6. 2026Self-Play Pretraining with Zero Data
  7. 2025Towards Understanding Self-play for LLM Reasoning
  8. 2026Self-Play Search Distillation for Large Language Model Reasoning