mlsanityData to 14 Oct 2026

Anchor-and-Harvest Curriculum: A Fully Unmasked Non-LLM Curriculum for Craftax

arXiv 2610.15328 ↗First posted 11 Oct 2026cs.LG cs.AIcited by 0

Not on the map. Why?

Abstract (arXiv)

We introduce Anchor-and-Harvest Curriculum (AHC), a domain-informed, rule-based training system for Craftax-Full. AHC combines achievement-oriented tasks, harvested initial-state progress, difficulty adaptation, bounded selection, and original-task practice. It uses no LLM inference and keeps all 43 actions available. At 2,005,401,600 PPO rollout transitions, three primary AHC runs obtain native-task return 46.70±0.57 (sample SD), compared with 41.91±1.01 for fresh native PPO-GTrXL with the same student architecture and evaluation protocol. Evaluation uses ten stochastic rollouts on each of 1024 fixed worlds per model. An additional cohort of three AHC runs reaches 46.08±1.23. Two single-component ablations retain similar mean returns, while jointly freezing difficulty and removing inherited progress gives 45.14±2.79. All-67-achievement profiles show that most of the return difference comes from Intermediate achievements, while the deepest-floor achievements remain unsolved. A Skill-Preflight diagnosis with a 14B teacher shows how prerequisite and floor-access constraints restrict practice. AHC and three 14B diagnostic configurations reach aggregate returns near the reported 80B DiCode mean of 45.88 and below the 235B mean of 48.55. These results show that an explicit curriculum without LLM inference can reach returns comparable to those reported for LLM-generated curricula on Craftax-Full, without action masking.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 23 of its 25 references that are in the map. First posted on arXiv 11 Oct 2026.

Builds on

23 of 25 references.

References · 23

  1. 2017Proximal Policy Optimization Algorithmscited by 15,215
  2. 2021Deep Reinforcement Learning at the Edge of the Statistical Precipicecited by 807
  3. 2017Deep Reinforcement Learning that Matterscited by 754
  4. 2020Emergent Complexity and Zero-shot Transfer via Unsupervised Environment Designcited by 275
  5. 2019Stabilizing Transformers for Reinforcement Learningcited by 270
  6. 2020First return, then explorecited by 266
  7. 2021Benchmarking the Spectrum of Agent Capabilitiescited by 200
  8. 2017Reverse Curriculum Generation for Reinforcement Learningcited by 198
  9. 2020Prioritized Level Replaycited by 198
  10. 2022Evolving Curricula with Regret-Based Environment Designcited by 173
  11. 2022Jump-Start Reinforcement Learningcited by 112
  12. 2021Replay-Guided Adversarial Environment Designcited by 111
11 more
  1. 2024Craftax: A Lightning-Fast Benchmark for Open-Ended Reinforcement Learningcited by 88
  2. 2018Learning Montezuma's Revenge from a Single Demonstrationcited by 80
  3. 2024OMNI-EPIC: Open-endedness via Models of human Notions of Interestingness with Environments Programmed in Codecited by 74
  4. 2023OMNI: Open-endedness via Models of human Notions of Interestingnesscited by 70
  5. 2023Empirical Design in Reinforcement Learningcited by 55
  6. 2024EnvGen: Generating and Adapting Environments via LLMs for Training Embodied Agentscited by 46
  7. 2024No Regrets: Investigating and Improving Regret Approximations for Curriculum Discoverycited by 31
  8. 2018Backplay: "Man muss immer umkehren"cited by 21
  9. 2024Refining Minimax Regret for Unsupervised Environment Designcited by 18
  10. 2026Dreaming in Code for Curriculum Learning in Open-Ended Worldscited by 3
  11. 2026Overcoming Valid Action Suppression in Unmasked Policy Gradient Algorithmscited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025Provable Benefit of Curriculum in Transformer Tree-Reasoning Post-Training
  2. 2026TAROT: Test-driven and Capability-adaptive Curriculum Reinforcement Fine-tuning for Code Generation with Large Language Models
  3. 2026LEACL: LLM-Enhanced Automatic Curriculum Learning for Reinforcement Learning in Long-Horizon Manipulation Tasks
  4. 2026Anticipating the Consequences of Curriculum Decisions with Large Language Models
  5. 2025AURA: Autonomous Upskilling with Retrieval-Augmented Agents
  6. 2025Self-Evolving Curriculum for LLM Reasoning
  7. 2026Unlocking the Unsolvable: Teacher-Guided Curriculum for Data-Efficient RLVR
  8. 2025EvoCurr: Self-evolving Curriculum with Behavior Code Generation for Complex Decision-making