Skip to content

Pull requests: areal-project/AReaL

New pull request New
78 Open 1,018 Closed
78 Open 1,018 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Reviews
Filter by reviews
No reviews Review required Approved review Changes requested
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

feat(workflow): add stage-1 stepwise reward plumbing
#1487 opened Jul 4, 2026 by Fyrgo8 • Draft
6 of 15 tasks
1
3
feat(infra): support Ray-managed HTTP proxy workers
#1486 opened Jul 4, 2026 by HughLLiu Loading…
6 of 15 tasks
8
fix(v2): close async clients on owner event loops
#1485 opened Jul 3, 2026 by jszzr Contributor Loading…
6 of 15 tasks
1
3
fix(v2): expose primary online rollout gateway
#1483 opened Jul 3, 2026 by jszzr Contributor • Draft
[codex] preserve LD_PRELOAD in local launchers
#1482 opened Jul 3, 2026 by jszzr Contributor • Draft
5
feat(rl): add version-checked online held-out evaluation
#1480 opened Jul 3, 2026 by jszzr Contributor • Draft
8 of 15 tasks
1
1
fix(v2): export local workflow statistics
#1478 opened Jul 3, 2026 by jszzr Contributor Loading…
8 of 15 tasks
1
fix(v2): separate callback and pull trajectory delivery
#1476 opened Jul 3, 2026 by jszzr Contributor Loading…
8 of 15 tasks
1
1
fix(v2): honor full-model disk weight updates
#1472 opened Jul 2, 2026 by jszzr Contributor Loading…
8 of 15 tasks
3
fix(rollout): stop controller-managed workers from dp-scaling staleness capacity
#1471 opened Jul 2, 2026 by Le8r0nJames Collaborator Loading…
4 of 15 tasks
3
fix(mcore): TP-shard GroupRMSNorm gate-norm weight for DCP checkpointing
#1470 opened Jul 2, 2026 by Le8r0nJames Collaborator Loading…
4 of 15 tasks
1
feat(distillation): Support cross-tokenizer on-policy distillation
#1452 opened Jun 30, 2026 by zahrayousefijamarani Contributor Loading…
4 of 13 tasks
1
6
feat(megatron): add MTP-augmented SFT/RL training
#1445 opened Jun 27, 2026 by HT-Yuan Collaborator • Draft
2 of 15 tasks
4
feat(vlm): add Qwen3.6 LoRA GRPO training support for 27B and 35B-A3B
#1444 opened Jun 26, 2026 by Lei00764 Loading…
6 of 15 tasks
10
feat(ppo): support actor loss aggregation modes
#1443 opened Jun 26, 2026 by EazyReal Contributor Loading…
8 of 15 tasks
5
feat(infra): add HTTP-based Ray Scheduler
#1441 opened Jun 26, 2026 by HwVanICI Collaborator Loading…
7 of 15 tasks
3
fix(io_struct): support multi-EOS models in stop-token handling
#1433 opened Jun 22, 2026 by PheelaV Loading…
8 of 15 tasks
8
fix(stats): reset single-key export metadata correctly
#1432 opened Jun 22, 2026 by EazyReal Contributor Loading…
4 tasks done
5
docs: mirgate and clean the documents
#1431 opened Jun 22, 2026 by mingcheng Contributor Loading…
8 of 15 tasks
3
feat(logging): add W&B worker GPU system metrics
#1428 opened Jun 21, 2026 by EazyReal Contributor Loading…
3 tasks done
3
fix(dataset): correct GSM8K SFT loss-mask boundary for merged tokens
#1427 opened Jun 21, 2026 by EazyReal Contributor Loading…
8 of 9 tasks
2
fix(reward): bound MathVerifyWorker.verify wall-clock on a hung verification
#1426 opened Jun 20, 2026 by EazyReal Contributor Loading…
1 task done
1
fix(api): normalize tokenizer-derived stop token ids
#1425 opened Jun 20, 2026 by EazyReal Contributor Loading…
7 of 9 tasks
5
refactor(workflow): extract grouped rollout wrapper stale
#1418 opened Jun 16, 2026 by RanranranQAQ Loading…
5 of 15 tasks
4
feat(ppo): add actor loss aggregation modes
#1417 opened Jun 16, 2026 by EazyReal Contributor Loading…
8 of 9 tasks
1
9
ProTip! Updated in the last three days: updated:>2026-07-02.