{"full_name":"rllm-org/rllm","owner":"rllm-org","name":"rllm","shard":"1","url":"https://github.com/rllm-org/rllm","description":"Democratizing Reinforcement Learning for LLMs","homepage":"https://docs.rllm-project.com","topics":["agent-framework","agentic-workflow","distributed-training","llm-training","machine-learning","ml-infrastructure","ml-platform","reinforcement-learning","coding-agent","llm-reasoning","search-agent","swe-agent","tinker","verl"],"tag_pages":["agent-framework","agentic-workflow","coding-agent","distributed-training","llm-training","machine-learning","reinforcement-learning"],"language":"Python","license":"Apache-2.0","avatar":"https://avatars.githubusercontent.com/u/221438497?v=4","created_at":"2025-01-26T05:31:40Z","pushed_at":"2026-10-16T19:02:56Z","metrics":{"stars":5919,"forks":643,"watchers":29,"issues":272,"prs":687,"releases":4,"issues_open":80,"issues_closed":192,"prs_open":126,"prs_closed":124,"prs_merged":437,"discussions":11},"growth":{"day":{"stars":1,"forks":-1,"base_day":"2026-11-15"}},"history":{"dates":["2026-11-10","2026-11-11","2026-11-12","2026-11-13","2026-11-14","2026-11-15","2026-11-16"],"stars":[5912,5916,5917,5917,5916,5918,5919],"forks":[638,642,643,643,643,644,643],"watchers":[null,30,29,29,29,29,29],"issues_open":[null,80,80,80,80,80,80],"prs_open":[null,104,124,124,124,126,126]},"boards":{},"same_owner":[],"issues":[{"number":30,"title":"oom when update_policy","comments":15,"created_at":"2025-02-15T15:51:14Z"},{"number":434,"title":"Qwen3.5 support","comments":13,"created_at":"2026-03-10T13:56:11Z"},{"number":29,"title":"Extremely Long Execution Time per Step","comments":12,"created_at":"2025-02-15T06:21:07Z"},{"number":33,"title":"Reproduce deepscaler","comments":10,"created_at":"2025-02-17T02:24:13Z"},{"number":7,"title":"great job！have some questions to ask","comments":10,"created_at":"2025-02-12T03:29:01Z"},{"number":67,"title":"Can we train the model on a 4 X 40G gpus or even fewer?","comments":9,"created_at":"2025-03-24T10:21:59Z"},{"number":12,"title":"Reproduce evaluation results on AIME (41% vs 43.1%)","comments":9,"created_at":"2025-02-12T08:37:07Z"},{"number":53,"title":"AssertionError: Expects sharded strategy","comments":8,"created_at":"2025-03-02T01:54:36Z"}],"discussions":[{"number":937,"title":"Research question: distinguishing failed evaluation from a genuine zero reward","comments":1,"category":"Ideas","created_at":"2026-10-17T20:03:45Z"},{"number":856,"title":"Toward a dependable rollout record in rLLM","comments":0,"category":"Ideas","created_at":"2026-08-14T16:42:08Z"},{"number":648,"title":"MCP to help OSS devs understand rllm-org/rllm","comments":0,"category":"General","created_at":"2026-06-12T12:07:57Z"},{"number":639,"title":"[Agentflow Refactor] How agent flows, sandboxes, tasks, and evaluators fit together","comments":0,"category":"Ideas","created_at":"2026-06-10T21:50:38Z"},{"number":492,"title":"Standardized rLLM Message Format: Analysis & Proposal","comments":0,"category":"Ideas","created_at":"2026-04-06T21:39:20Z"},{"number":251,"title":"rLLM v0.2 release","comments":0,"category":"Announcements","created_at":"2025-10-16T21:24:26Z"},{"number":231,"title":"rllm Container Image size too big","comments":0,"category":"General","created_at":"2025-10-02T21:09:46Z"},{"number":159,"title":"Is there a groundtruth solution available for all the deepcoder datasets?","comments":2,"category":"Q&A","created_at":"2025-07-23T17:03:34Z"}],"latest_release":{"name":"rLLM: v0.3.0-pre","tag":"v0.3.0-pre","published_at":"2026-04-30T07:22:47Z","body":"## What's Changed\r\n* [feature] add support for on policy distillation by @kylemontgomery1 in https://github.com/rllm-org/rllm/pull/356\r\n* [Bugfix] Make sure the tracking logger is now explicitly finished by trainers by @listar2000 in https://github.com/rllm-org/rllm/pull/358\r\n* Add Geo3K Tinker training example with VLM support by @BrianChen26 in https://github.com/rllm-org/rllm/pull/357\r\n* Support custom metric for SDK by @thwu1 in https://github.com/rllm-org/rllm/pull/362\r\n* fix: typo in bibtex by @tongyx361 in https://github.com/rllm-org/rllm/pull/363\r\n* Fireworks API returns raw token ids to avoid re-tokenization by @1stprinciple in https://github.com/rllm-org/rllm/pull/301\r\n* overlong_filter in stepwise RL by @LianShuQuan in https://github.com/rllm-org/rllm/pull/365\r\n* Verifiers and Prime Intellect Environment Integration & Example by @alt-glitch in https://github.com/rllm-org/rllm/pull/367\r\n* Add test_train.sh for docs build by @alt-glitch in https://github.com/rllm-org/rllm/pull/374\r\n* Resolve Tinker error in advantage computation by @listar2000 in https://github.com/rllm-org/rllm/pull/375\r\n* Merge nightly branch into main by @listar2000 in https://github.com/rllm-org/rllm/pull/377\r\n* [Fix] Add a few patches to unified trainer by @listar2000 in https://github.com/rllm-org/rllm/pull/383\r\n* [Doc improvements] Add the missing multi-GPU configuration options to the evaluation scripts in the quick-start documentation. by @SyncLionPaw in https://github.com/rllm-org/rllm/pull/384\r\n* [Fix] Further enhancements to the Unified Trainer by @listar2000 in https://github.com/rllm-org/rllm/pull/385\r\n* [Experimental] Fixing Tinker rollout prompt ids by @listar2000 in https://github.com/rllm-org/rllm/pull/390\r\n* Reset workflow with task and uid before workflow execution by @JasonWei05 in https://github.com/rllm-org/rllm/pull/389\r\n* Fully Async Trainer by @thwu1 in https://github.com/rllm-org/rllm/pull/394\r\n* Add LLM-in-Sandbox to awesome projects by @cdxeve in https://github.com/rllm-org/rllm/pull/395\r\n* 【Fix】All content in examples/search were empty by @Fulin-Gao in https://github.com/rllm-org/rllm/pull/396\r\n* [Feat] Important updates to the experimental unified trainer by @listar2000 in https://github.com/rllm-org/rllm/pull/398\r\n* FinQA Release by @mananroongta in https://github.com/rllm-org/rllm/pull/393\r\n* Refactor unified trainer async flow and add precomputed-advantage + Tinker rollout fixes by @listar2000 in https://github.com/rllm-org/rllm/pull/401\r\n* feat(tracking): integrate changes need for the rllm-ui by @Chanbinski in https://github.com/rllm-org/rllm/pull/402\r\n* Fixed the qwen parser to handle case when model is non-thinking but user put disable_thinking=false as config by @jeewoo-lee in https://github.com/rllm-org/rllm/pull/405\r\n* Enable TrajectorygGoup's role specific RL advantage estimator via unified trainer hook by @listar2000 in https://github.com/rllm-org/rllm/pull/408\r\n* Move Tinker backend to rllm.trainer.tinker and deprecate legacy trainer APIs under rllm.trainer.deprecated by @listar2000 in https://github.com/rllm-org/rllm/pull/409\r\n* Refactor RL advantage estimators and add REINFORCE++ baseline/RLOO support by @listar2000 in https://github.com/rllm-org/rllm/pull/410\r\n* Extend On-Policy Distillation support with unified trainer example by @BrianChen26 in https://github.com/rllm-org/rllm/pull/406\r\n* [Doc] Unified trainer docs by @listar2000 in https://github.com/rllm-org/rllm/pull/414\r\n* [Fix] Put doc images into assets by @listar2000 in https://github.com/rllm-org/rllm/pull/415\r\n* [Fix] Hot-fix advantage calcuilation unpacking issue by @listar2000 in https://github.com/rllm-org/rllm/pull/416\r\n* Add deprecation notices in docs by @listar2000 in https://github.com/rllm-org/rllm/pull/417\r\n* feat(tracking): UILogger with non-blocking background worker by @Chanbinski in https://github.com/rllm-org/rllm/pull/419\r\n* feat(rllm-model-gateway): add standalone proxy server for rLLM  by @luyuzhe111 in https://github.com/rllm","is_prerelease":false},"readme":true,"readme_zh":false,"news":[],"stale":false,"snapshot_at":"2026-11-16T09:23:14Z","history_start":"2026-09-06"}