CogEvol: Towards Efficient and Reliable Learning Environment Generation
Not on the map. Why?
Abstract (arXiv)
We present CogEvol, a family of models trained specifically for Learning Environment Generation: turning a course brief into a finished learning artifact (structured-JSON slides or self-contained interactive HTML pages) in a single pass. Across 220k production requests, CogEvol completes a slide in a median of 17 seconds and an interactive page in 59, replacing minutes-long multi-turn agent scaffolding. Reliability is enforced rather than hoped for: a production-grounded data pipeline turns real failures into 53,687 verified SFT samples, and a hybrid rule-plus-VLM reward drives GRPO-based RL, hardened after we caught and fixed a reward-hacking episode that produced visually convincing but unplayable games. CogEvol-27B scores 83.7 on slide quality and 63.7 on a 500-case interactive-HTML benchmark with 26.9x fewer parameters than flagship coding models, and, in collaboration with the OpenMAIC team, serves their live production traffic. CogEvol-4B is released openly under the Apache 2.0 license at https://github.com/CogEvol/CogEvol-4B; external flagships are measured on the same suites under the identical harness. Scaffold editing cuts interactive-page generation cost by a further ~76%, and the full stack runs on domestic Ascend accelerators at application-level parity with A800 GPUs, lowering the unit cost of AI-native education at scale.
Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 17 of its 33 references that are in the map. First posted on arXiv 31 Aug 2026.
Builds on
17 of 33 references.
References · 17
- 2023Efficient Memory Management for Large Language Model Serving with PagedAttentioncited by 5,767
- 2022FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awarenesscited by 3,611
- 2025DAPO: An Open-Source LLM Reinforcement Learning System at Scalecited by 2,678
- 2019Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelismcited by 1,829
- 2017Quantization and Training of Neural Networks for Efficient Integer-Arithmetic-Only Inferencecited by 1,548
- 2024OpenHands: An Open Platform for AI Software Developers as Generalist Agentscited by 956
- 2025DeepSeek-V3.2: Pushing the Frontier of Open Large Language Modelscited by 755
- 2026GLM-5: from Vibe Coding to Agentic Engineeringcited by 502
- 2024Gated Delta Networks: Improving Mamba2 with Delta Rulecited by 483
- 2024Better & Faster Large Language Models via Multi-token Predictioncited by 344
- 2024Large Language Models for Education: A Survey and Outlookcited by 185
- 2024LLM-Based Test-Driven Interactive Code Generation: User Study and Empirical Evaluationcited by 46
5 more
- 2026PresentBench: A Fine-Grained Rubric-Based Benchmark for Slide Generationcited by 12
- 2023LittleMu: Deploying an Online Virtual Teaching Assistant via Heterogeneous Sources Integration and Chain of Teach Promptscited by 9
- 2024New Solutions on LLM Acceleration, Optimization, and Applicationcited by 7
- 2019How Different Are Different diff Algorithms in Git?cited by 6
- 2026Evaluating Interactivity: Toward Automated Assessment of AI-Generated Explorable Explanationscited by 1
Built on by
Not cited yet.
Similar, not cited
Close in wording, no citation either way
- 2025AutoEnv: Automated Environments for Measuring Cross-Environment Agent Learning
- 2025GenEnv: Difficulty-Aligned Co-Evolution Between LLM Agents and Environment Simulators
- 2025Fine-tuning with RAG for Improving LLM Learning of New Skills
- 2026GenEvolve: Self-Evolving Image Generation Agents via Tool-Orchestrated Visual Experience Distillation
- 2022GriddlyJS: A Web IDE for Reinforcement Learning
- 2026A Deep Dive into Scaling RL for Code Generation with Synthetic Data and Curricula
- 2026Co-Evolving LLM Decision and Skill Bank Agents for Long-Horizon Tasks
- 2026Learning to Present: Inverse Specification Rewards for Agentic Slide Generation