[Collaboration] Benchmark your repair loop on LoopBench (LB-CR-1)

[KanakMalpani]

Collaboration invite — benchmark your repair loop on LoopBench

Hi @OpenAutoCoder maintainers — Agentless solves software repair with a closed locate → patch → verify loop, which maps cleanly to our public LB-CR-1 (code repair) task.

We run a fixed-seed, no-API-key scoreboard for comparable loop engineering:

Live board	https://kanakmalpani.github.io/LoopBench/
Task	LB-CR-1 — fix broken code under test-suite pressure
60-second path	https://github.com/KanakMalpani/Loop-Engineering/blob/main/contributions/BEAT_LB-CR-1.md
Row schema	https://github.com/KanakMalpani/LoopBench/blob/main/leaderboard/ROW_SCHEMA.md

pip install "le-loopforge>=0.2.0" "le-loopctl>=0.1.0" "loopbench>=0.1.1" "loopgym>=0.1.2"
loopbench run --task LB-CR-1 --spec your-agentless-loop.yaml --seeds 0,1,2,3,4 -o results.json
loopbench validate results.json
# → PR to KanakMalpani/LoopBench leaderboard/entries.json

What we need: one maintainer or contributor PR with your loop mapped to LSS YAML + observed LES. First non-maintainer row gets permanent leaderboard credit and a shout-out on Loop Engineering.

Happy to pair on the LSS mapping (workers / evaluators / termination_conditions) if useful. Playground hub: https://github.com/KanakMalpani/Loop-Engineering/blob/main/contributions/LOOP_PLAYGROUND.md

— Loop Engineering community (issue #4)

Posted 2026-06-25 UTC via adoption_wave11.py

[KanakMalpani]

Follow-up — partner submission pack (30 min)

Thanks again for considering a LoopBench row. We shipped a fork-and-PR kit since the initial invite:

Partner guide: https://github.com/KanakMalpani/Loop-Engineering/blob/main/contributions/PARTNER_LOOPBENCH_SUBMIT.md

pip install "le-loop-stack[bench]>=0.1.0"
git clone https://github.com/KanakMalpani/Loop-Engineering.git
cd Loop-Engineering
python scripts/run_submission_dryrun.py --partner agentless   # or aider / openhands
loopbench run --task LB-CR-1 --spec docs/submission-dry-run/partner-*-lb-cr-1.yaml --seeds 0,1,2,3,4 -o results.json
loopbench validate results.json
# → Fork LoopBench, add row to leaderboard/entries.json, open PR

• BEAT path: https://github.com/KanakMalpani/Loop-Engineering/blob/main/contributions/BEAT_LB-CR-1.md
• Live board: https://kanakmalpani.github.io/LoopBench/
• Maintainer pairing offered — reply here or on Loop-Engineering #4

First non-maintainer merge gets permanent leaderboard credit.

Posted 2026-06-25 UTC via adoption_wave12.py