Issues

• [RFC] Define the generated-task feedback state and keep it in the recipe

  enhancementRFC
  Human-Agent-Society/reef#912 · Filter by author BobbyZhouZijian opened on Oct 24, 2026

• [Bug] Reporting an inference record without policy tensors stops automatic training

  area: corebugstatus: needs-triage
  Human-Agent-Society/reef#909 · Filter by author augustuszzq opened on Oct 23, 2026

• [RFC] Submit bounded training jobs to training components instead of driving each step

  area: servicearea: trainingenhancementRFCstatus: needs-triage
  Human-Agent-Society/reef#899 · Filter by author BobbyZhouZijian opened on Oct 23, 2026

• [Experiment] Run ServeLearnBench through Reef and train on it with SAO and OPD

  area: examplesarea: trainingstatus: needs-triagetype: experiment
  Human-Agent-Society/reef#893 · Filter by author MikeZheng777 opened on Oct 23, 2026
  1

  

• [RFC] Separate the serving endpoint from the training rollout engine

  area: servicearea: trainingenhancementRFCstatus: needs-triage
  Human-Agent-Society/reef#889 · Filter by author BobbyZhouZijian opened on Oct 22, 2026
  1

• [RFC] Keep SQLite commits in the scenario database

  enhancementRFCstatus: ready
  Human-Agent-Society/reef#887 · Filter by author BobbyZhouZijian opened on Oct 22, 2026
  1

• [RFC] One task store of Harbor directories: import operator tasks, and let Cordis evaluate on its eval split

  enhancementRFCstatus: ready
  Human-Agent-Society/reef#870 · Filter by author BobbyZhouZijian opened on Oct 20, 2026

• [Feature] Record2batch 2/3: Scheduling, generation and buffering

  area: trainingenhancement
  Human-Agent-Society/reef#832 · Filter by author simonucl opened on Oct 16, 2026

• [Feature] Record2batch 3/3: Rollout and training batches

  area: trainingenhancement
  Human-Agent-Society/reef#833 · Filter by author simonucl opened on Oct 16, 2026

• [Feature] Keep agent episodes with context compaction usable for training and task generation

  Human-Agent-Society/reef#825 · Filter by author simonucl opened on Oct 15, 2026
  2

• [Community] Reef office hour agendas and minutes

  roadmapstatus: keep-opentype: task
  Human-Agent-Society/reef#818 · Filter by author BobbyZhouZijian opened on Oct 15, 2026
  2

• [RFC] Add ArenaRL pairwise ranking to weight training

  enhancementRFCstatus: needs-triage
  Human-Agent-Society/reef#812 · Filter by author BobbyZhouZijian opened on Oct 14, 2026
  

• [RFC] Adapt record2task generation from solver feedback

  area: trainingenhancementRFCstatus: ready
  Human-Agent-Society/reef#811 · Filter by author BobbyZhouZijian opened on Oct 14, 2026
  

• [RFC] Report resource usage for each scenario

  area: observabilityenhancementpriority: p3RFCstatus: ready
  Human-Agent-Society/reef#810 · Filter by author BobbyZhouZijian opened on Oct 14, 2026

• [RFC] Generated tasks as record2task output, with rollout in task2batch

  area: trainingenhancementRFCstatus: ready
  Human-Agent-Society/reef#805 · Filter by author BobbyZhouZijian opened on Oct 14, 2026
  2

• [RFC] Declare tree node checks with matchers in harness descriptors

  area: harnessenhancementRFCstatus: needs-triage
  Human-Agent-Society/reef#797 · Filter by author yanxz opened on Oct 14, 2026
  1

  

• [Task] Split the "Harness Adapter" page

  area: docsstatus: readytype: task
  Human-Agent-Society/reef#788 · Filter by author yanxz opened on Oct 13, 2026
  1

  

• [RFC] One worker concept for background jobs and evaluation workers

  area: trainingenhancementRFCstatus: ready
  Human-Agent-Society/reef#787 · Filter by author BobbyZhouZijian opened on Oct 13, 2026

• [RFC] Record evaluation inputs and score saved runs again

  Human-Agent-Society/reef#770 · Filter by author shanshan1017 opened on Oct 12, 2026
  2

• [Community] Multimodal Roadmap (Draft)

  area: corearea: trainingroadmap
  Human-Agent-Society/reef#749 · Filter by author quao627 opened on Oct 10, 2026
  10

• [RFC] Run each Reef component as a separately managed service

  area: packagingenhancementRFCstatus: needs-triage
  Human-Agent-Society/reef#748 · Filter by author BobbyZhouZijian opened on Oct 10, 2026
  4

• [RFC] Run more than one Reef service instance by giving each scenario one owner

  enhancementRFCstatus: needs-triage
  Human-Agent-Society/reef#747 · Filter by author BobbyZhouZijian opened on Oct 10, 2026

• [RFC] Integrate Semantic Router for Reef model routing

  area: serviceenhancementRFCstatus: ready
  Human-Agent-Society/reef#741 · Filter by author RickWayne1125 opened on Oct 9, 2026
  6

  

• [Experiment] sao: add sync feat and test it

  area: trainingstatus: needs-triagetype: experiment
  Human-Agent-Society/reef#735 · Filter by author yanan1116 opened on Oct 8, 2026
  1

• [RFC] Support a larger same-tokenizer teacher for OPD

  area: trainingenhancementRFCstatus: ready
  Human-Agent-Society/reef#701 · Filter by author MikeZheng777 opened on Oct 1, 2026
  1