Show HN: A Lego-like composable Framework for RSI

Researchers from Imperial College London, Duke University, the Chinese University of Hong Kong, Brown University and an independent researcher released GEAR, an open framework that organizes recursive self-improvement of LLM agent harnesses as a two-level meta-learning process with an inner rollout loop and an outer Meta Agent loop. In same-model comparisons, GEAR's DSH harness raised task pass rates on AutomationBench Marketing from 27.0% to 40.0%, on AutomationBench Support from 17.0% to 25.0%, and on Terminal-Bench 2.1 from 52.8% to 62.9% with GPT 5.6 Luna at medium reasoning effort, and improved GPT 6 Astra max results over Codex by 4.0, 4.0 and 0.9 percentage points respectively. Code and documentation are published on GitHub at rsi-gear/gear and at rsigear.xyz/docs/gear.

GEAR: An Open General Evolutionary Architecture for Recursive Self-Improvement of LLM Agent Harness <sup 1</sup Imperial College London <sup 2, 5</sup Independent Researcher <sup 3</sup Duke University <sup 4</sup The Chinese University of Hong Kong <sup 6</sup Brown University Learn from experience. Improve across tasks. GEAR organizes recursive self-improvement as a two-level meta-learning process. In the inner loop , a rollout agent uses task feedback to refine its reasoning, planning, and actions within a task. In the outer loop , a Meta Agent identifies common patterns across task experiences, including recurring failures and effective strategies, and proposes reusable improvements to enhance cross-task performance. GEAR evaluates and promotes these changes for subsequent tasks. The two loops can be optimized independently and are connected by experience. Modular algorithms for evolutionary search. GEAR organizes evolutionary search into replaceable algorithm components, providing a common framework for studying different search strategies. Staged GEPA is one concrete instantiation, illustrated below. Open the search, component by component. Parent selection Choose which archived harness to improve next. Staged GEPA uses task-level fronts to retain useful specialists and sample parents from the frozen archive. - Reads - Frozen archive, champion, task-level evidence - Produces - A parent harness for candidate generation Policy modules express the framework’s design boundaries. The current staged implementation exposes parent selection as a public policy hook; evidence validation, evaluation stages, and final promotion remain controlled by the runtime. Better harnesses. More completed tasks. Improvements across coding, marketing, and support. Same-model comparisons, with reasoning effort made explicit. Marketing · task pass rate27% → 40%GPT 5.6 Luna · medium · DSH Base → OursSupport · task pass rate17% → 25%GPT 5.6 Luna · medium · DSH Base → OursCoding · task pass rate52.8% → 62.9%GPT 5.6 Luna · medium · DSH Base → Ours Full benchmark comparison | Harness comparison · same model and reasoning effort within each group | | | | | | | |---|---|---|---|---|---|---| | Agent | Harness | AutomationBench · Marketing ↑ | | AutomationBench · Support ↑ | | Terminal-Bench 2.1 ↑ | |---|---|---|---|---|---|---| | | | Task pass rate | Obj. completion | Task pass rate | Obj. completion | Task pass rate | | GPT 5.6 Luna medium | DSH Base | 27.0% | 75.4% | 17.0% | 72.9% | 52.8% | | | DSH Ours | 40.0% | 83.9% | 25.0% | 81.3% | 62.9% | | | Δ pp | +13.0 | +8.5 | +8.0 | +8.4 | +10.1 | | GPT 6 Astra max | Codex | 57.0% | 84.1% | 35.0% | 88.4% | 86.7% | | | DSH Ours | 61.0% | 86.3% | 39.0% | 88.7% | 87.6% | | | Δ pp | +4.0 | +2.2 | +4.0 | +0.3 | +0.9 | Δ is the absolute improvement in percentage points. | Additional evaluation · GPT 5.6 Luna max · DSH Ours | | | |---|---|---| | Benchmark | Task pass rate | Objective completion | |---|---|---| | AutomationBench · Marketing | 53.0% | 88.9% | | AutomationBench · Support | 27.0% | 82.4% | | Terminal-Bench 2.1 | 84.3% | — | AutomationBench reports task pass rate and objective completion; Terminal-Bench 2.1 reports task pass rate. Luna max is a separate evaluation at a higher reasoning effort. Result data ↗ assets/paper-results.json Optimize. Execute. Understand. Separate responsibilities, connected by versioned evidence. Explore GEAR. Code, experiment records, and the tools to start your own research. 01 / Source ↗ https://github.com/rsi-gear/gear Get the code Explore the framework, search components, and examples. github.com/rsi-gear/gear 02 / Documentation ↗ https://rsigear.xyz/docs/gear Run your first experiment Set up a task benchmark and connect the Refine Skill. Read the guide 03 / Experiment ↗ https://github.com/rsi-gear/gear/blob/main/docs/guide/en/example-algorithm.md Follow the search Inspect the staged Marketing experiment and its evaluation evidence. View the experiment Cite GEAR. Preprint · October 2026. arXiv link will be added when available. @unpublished{tang2026gear, title = {{GEAR}: An Open General Evolutionary Architecture for Recursive Self-Improvement of {LLM} Agent Harness}, author = {Tang, Yehui and Zhu, Guanqi and Xia, Shao-Jun and Chen, Xin and Geng, Mengzhe and Du, Hongyang and Tang, Yuhao}, year = {2026}, note = {Preprint}, url = {https://github.com/rsi-gear/gear} }