Skip to content

Pull requests: PrimeIntellect-ai/prime-rl

New pull request New
71 Open 2,101 Closed
71 Open 2,101 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

[codex] Add experimental Dynamo backend
#2472 opened May 11, 2026 by samsja Member • Draft
feat: opt-in renderer-based tokenization for SFT
#2471 opened May 10, 2026 by ercbot Loading…
1
2
Fix: add renderers>=0.1.5 as direct dependency and bump verifiers rev
#2469 opened May 10, 2026 by ercbot Loading…
1
feat: json logging for inference (vLLM)
#2466 opened May 10, 2026 by JannikSt Member • Draft
13
feat: add Dynamo vLLM inference backend
#2465 opened May 10, 2026 by samsja Member • Draft
[Weight Transfer]: Support layerwise reloading for online quantization
#2464 opened May 9, 2026 by S1ro1 Collaborator Loading…
3
[codex] add sglang backend and nccl broadcast
#2462 opened May 9, 2026 by samsja Member • Draft
feat(trainer): symmetric DPPO + KL on unmasked tokens (vLLM 0.19 base)
#2460 opened May 9, 2026 by samsja Member • Draft
feat(orchestrator): drop groups after max_error_reschedule_attempts
#2459 opened May 9, 2026 by samsja Member Loading…
2
feat(trainer): KL penalty over unmasked tokens only
#2458 opened May 9, 2026 by samsja Member • Draft
pin images to cudnn cu128. this shold break ci because of the vllm de…
#2451 opened May 8, 2026 by Jackmin801 Member Loading…
[codex] Add synchronous ES LoRA trainer
#2447 opened May 8, 2026 by Matthew-agi • Draft
[DYNAMO] wire TITO client to Dynamo nvext transport
#2446 opened May 8, 2026 by AmeenP Contributor • Draft
[DYNAMO] smoke runner follow-up from tested branch
#2445 opened May 8, 2026 by AmeenP Contributor • Draft
[DYNAMO] fix: guard lazy_load_kernel compat patch
#2444 opened May 8, 2026 by AmeenP Contributor • Draft
fix(inference): RoutedExpertsCapturer for DeepEP MK path + cudagraph …
#2440 opened May 8, 2026 by samsja Member • Draft
feat(inference): fp32 lm_head for stable logprobs under FP8/bf16
#2438 opened May 7, 2026 by samsja Member Loading…
4
feat(renderer): expose preserve_*_thinking flags in RendererConfig
#2436 opened May 7, 2026 by hallerite Member • Draft
feat(trainer): symmetric DPPO-Binary TV default loss (no KL, no advantage conditioning)
#2434 opened May 7, 2026 by samsja Member • Draft
feat(glm_moe_dsa): experimental bucketed sparse-MLA fwd/bwd kernels
#2432 opened May 6, 2026 by S1ro1 Collaborator • Draft
fix(monitor): durable sample-upload retries with bounded backlog
#2425 opened May 5, 2026 by JannikSt Member Loading…
19
feat(trainer): IPO+KL — double-sided mask, KL only on kept tokens
#2423 opened May 5, 2026 by samsja Member • Draft
chore: default weight_decay to 0 in BaseOptimizerConfig
#2422 opened May 5, 2026 by samsja Member Loading…
[DYNAMO] pass renderer transport to verifiers
#2420 opened May 5, 2026 by AmeenP Contributor • Draft
feat(k8s): minimal kubectl launcher mirroring slurm
#2419 opened May 5, 2026 by samsja Member • Draft
ProTip! What’s not been updated in a month: updated:<2026-04-11.