Skip to content

AgentFlow: In-the-Flow Agentic System Optimization

active 2025-09-272026-05-30 (UTC)

Partial coverage11,280 / 12,041 hourly files (94%) · 2 absent upstream · 758 failed, retryable2025-03-272026-08-10 (UTC)— sampled evenly across the window, so rankings and trends hold; absolute counts scale up.
Events
491
Pushes
26
Pull requests
1
Issues
20
Stars
397
Forks
14

Activity over time

Daily event counts in the loaded window

Line chart, 246 days from 2025-09-27 to 2026-05-30. Pushes: 26 total, peak 6 in a day. Pull requests: 1 total, peak 1 in a day. Issues: 20 total, peak 2 in a day. Comments: 30 total, peak 4 in a day. Stars: 397 total, peak 16 in a day.

  • Pushes
  • Pull requests
  • Issues
  • Comments
  • Stars

Stars, PRs, issues and forks are under-captured in the later part of this window. GH Archive progressively stopped capturing non-push events during 2026 — −95% or worse by the end of the window. Every series here except Pushes fades for that reason, so a decline above reflects the archive, not this repository. Pushes stay reliable throughout, so read them, and the contributor counts derived from them, as the real signal. Data health has the measurements.

Top contributors

Pushes, PRs, issues, reviews and comments — stars and forks excluded, so this is contribution rather than popularity

ContributorContributionsPushesPRsComments
IsaacGHX3420010
Zhuofeng-Li20309
Curry30Messi4004
SimonHeye4001
lupantech4300
mgh11001
ymyamz1001
Nardien1000
yanyiwei1001
han678901000
devrajeev12341010
lebronjamesking1001
xingyuHuxingyu1001
arghavan-kpm1000
HelloWorldLTY1000
zhangxixi5511001

Recent activity

Latest issues, pull requests and releases

  • Issue#39han678902026-04-26 13:36
    Question about API
  • Issue#38Zhuofeng-Li2026-03-11 00:54
    Typo in project page
  • Issue comment#38Zhuofeng-Li2026-03-11 00:54
    Typo in project page
  • Issue#36SimonHeye2026-03-07 10:45
    An inquiry about the experimental results
  • Issue comment#33IsaacGHX2026-01-30 08:24
    How to run AgentFlow Benchmark with a locally trained Flow-GRPO checkpoint
  • Issue comment#8Zhuofeng-Li2026-01-27 01:54
    A request for guidance on a comparative experiment
  • Issue#24IsaacGHX2026-01-27 01:45
    WandB Training Logs
  • Issue#21IsaacGHX2026-01-27 01:36
    Can't Reproduce AIME24 Results following readme.
  • Issue comment#21IsaacGHX2026-01-27 01:36
    Can't Reproduce AIME24 Results following readme.
  • Issue#32arghavan-kpm2026-01-24 23:53
    Training cost
  • Issue comment#28yanyiwei2026-01-24 01:40
    Has anyone actually reproduced the results of this repo?
  • Issue comment#30IsaacGHX2026-01-23 10:37
    Signal Module Error in PythonCoderTool During RL Training
  • Issue comment#23IsaacGHX2026-01-23 10:09
    Unable to reproduce the AgentFlow experimental results on AIME24.
  • Issue comment#24Zhuofeng-Li2026-01-23 03:05
    WandB Training Logs
  • Issue comment#30Zhuofeng-Li2026-01-23 03:04
    Signal Module Error in PythonCoderTool During RL Training
  • Issue#25IsaacGHX2026-01-22 05:46
    Automatic vLLM service launch during Flow-GRPO training
  • Issue#27lupantech2026-01-22 05:45
    Concerns on Experimental Validity: External LLM Dependency in Google Search Tool and Data Leakage in Open-Source Benchmarks
  • Issue comment#30IsaacGHX2026-01-22 04:54
    Signal Module Error in PythonCoderTool During RL Training
  • Issue comment#23xingyuHuxingyu2025-12-21 04:53
    Unable to reproduce the AgentFlow experimental results on AIME24.
  • Issue#17Zhuofeng-Li2025-12-16 02:50
    Memory and Google Search do not handle multilines
  • Issue#20Zhuofeng-Li2025-12-16 02:50
    The wechat_group QR code has expired.
  • Issue comment#22ymyamz2025-12-12 15:25
    Plan Module Training Deployment Issue
  • Issue comment#22zhangxixi5512025-12-10 09:06
    Plan Module Training Deployment Issue
  • Issue comment#19IsaacGHX2025-12-09 02:57
    [question] What magnitude of training data do we need to use AgentFlow?
  • Issue comment#7IsaacGHX2025-11-20 08:06
    During inference, how to replace openai model for judging response with a local vllm served model

Totals cover only the window loaded into ClickHouse and count events, not GitHub's lifetime totals — 397 stars here means stars gained during the window, not the repo's star count.