terminal-bench-mini: 14 tasks reproducing the Terminal-Bench 2.0 ranking

README.md

@@ -0,0 +1,139 @@
+---
+license: apache-2.0
+task_categories:
+- text-generation
+tags:
+- benchmark
+- agents
+- terminal-bench
+- evaluation
+size_categories:
+- n<1K
+configs:
+- config_name: tasks
+  data_files: tasks.csv
+- config_name: leaderboard
+  data_files: leaderboard.csv
+- config_name: mini_trials
+  data_files: mini_trials.csv
+- config_name: size_curve
+  data_files: size_curve.csv
+---
+
+# terminal-bench-mini
+
+Fourteen of Terminal-Bench 2.0's ninety tasks, picked so that ranking agents on
+the subset reproduces ranking them on the whole benchmark.
+
+Running ninety tasks five times each is how the official leaderboard is built.
+That is out of reach if you are comparing quant variants, fine-tunes or local
+models on your own hardware. This subset turns a multi-day sweep into a few
+hours.
+
+Same approach as [deepswe-mini](https://huggingface.co/datasets/LocalLLaMA/deepswe-mini):
+take the published per-task results, rank the field on the full benchmark, then
+find a small subset that preserves that ranking. **These fourteen tasks
+reproduce about 87% of the full benchmark's pairwise ordering, and 97% of it
+when two agents are more than ten points apart.**
+
+## The tasks
+
+| task | field pass rate | spread across agents | run-to-run noise |
+|---|---|---|---|
+| filter-js-from-html | 11.9% | 0.310 | 0.022 |
+| install-windows-3.11 | 11.9% | 0.258 | 0.090 |
+| mteb-retrieve | 41.2% | 0.420 | 0.149 |
+| db-wal-recovery | 56.8% | 0.448 | 0.097 |
+| qemu-alpine-ssh | 69.3% | 0.417 | 0.093 |
+| build-pmars | 75.2% | 0.361 | 0.122 |
+| mcmc-sampling-stan | 81.5% | 0.328 | 0.100 |
+| rstan-to-pystan | 81.7% | 0.329 | 0.102 |
+| kv-store-grpc | 86.4% | 0.297 | 0.063 |
+| bn-fit-modify | 87.6% | 0.295 | 0.056 |
+| regex-log | 90.8% | 0.270 | 0.025 |
+| custom-memory-heap-crash | 91.9% | 0.273 | 0.000 |
+| crack-7z-hash | 93.5% | 0.228 | 0.022 |
+| vulnerable-secret | 94.6% | 0.226 | 0.000 |
+
+Field pass rate is the mean over 37 agents. Spread is the standard deviation of
+those per-agent rates: a task with spread near zero tells you nothing about who
+is better. Run-to-run noise is the average standard deviation across repeated
+trials of the same agent on the same task.
+
+Task definitions live in Terminal-Bench 2.0. This dataset names them, it does
+not redistribute them.
+
+## Method
+
+The source is every trial published to the Terminal-Bench 2.0 leaderboard:
+**32,803 trials, 27,405 of them scorable**, across 90 tasks and 75 agent/model
+submissions. Trials whose environment failed to start carry `exception_info` and
+no verifier result; they are dropped rather than counted as failures. The 37
+submissions that attempted at least 85 tasks form the reference ranking.
+
+The subset is a stratified sample. Tasks that every agent passes or every agent
+fails carry no ranking information and are excluded first. The rest are sorted
+by field pass rate into seven difficulty bands; within each band the quieter
+half is preferred, since a task that flips between identical runs costs every
+future user extra trials to average out; two tasks are drawn per band with a
+fixed seed (20260919).
+
+Quality is measured as pairwise agreement: over every pair of agents, how often
+the subset orders them the way all ninety tasks do, broken out by how far apart
+the pair really is.
+
+| pair separation | agreement |
+|---|---|
+| more than 10 points | 97.3% |
+| 5 to 10 points | 80.1% |
+| 2 to 5 points | 73.2% |
+| under 2 points | 65.1% |
+
+Agreement also rises with how many tasks you run, which `size_curve.csv`
+records: 14 tasks average 85%, 22 average 89%, 30 average 91%, 60 average 95%.
+If the systems you are comparing are close, add tasks or add trials.
+
+## What it is good for
+
+Use it to answer "is this configuration roughly competitive". The top of the
+table is stable — the leading agent leads on both, and every agent in the full
+top ten stays in the mini top twelve. Maximum rank movement is 16 places, and it
+happens in the crowded middle where full-benchmark scores differ by fractions of
+a point.
+
+Do not use it to claim one agent beats another by two points. Five of the
+fourteen tasks have run-to-run noise above 0.09, so run more than one trial per
+task or expect a few points of movement that mean nothing.
+
+## Files
+
+- `tasks.csv` — the fourteen tasks with their field statistics
+- `leaderboard.csv` — all 37 agents scored on the mini set and the full 90, with both ranks and the movement between them
+- `mini_trials.csv` — the agent × task pass rates behind that leaderboard
+- `size_curve.csv` — agreement against subset size, 200 random subsets per size
+
+## Reproducing
+
+Source data is the public `harborframework/terminal-bench-2-leaderboard`
+repository. Every trial record is a small `result.json` sitting next to terminal
+logs that make the repository 121 GB, so fetch only the records:
+
+```bash
+git clone --filter=blob:none --no-checkout \
+  https://huggingface.co/datasets/harborframework/terminal-bench-2-leaderboard
+cd terminal-bench-2-leaderboard
+git sparse-checkout set --no-cone '/**/result.json'
+git checkout
+```
+
+That is 348 MB and about two minutes. Each record carries
+`verifier_result.rewards.reward`. Note that two task-name conventions coexist in
+the repository, `terminal-bench/<name>` and `<name>`; normalise them or the same
+task counts twice and no two agents ever overlap.
+
+## Caveats
+
+Agreement is measured against Terminal-Bench 2.0 as scored by these 37
+submissions, all of them frontier hosted models. A small local model sits below
+that range, where the subset is untested — though a gap that large is the case
+it handles most reliably.

leaderboard.csv

@@ -0,0 +1,38 @@
+rank_mini_14,rank_full_90,rank_delta,agent,mini_14,full_90,tasks_attempted
+1,1,0,vix (claude-opus-4-7),1.0,0.9971,85
+2,3,-1,logos-latest (claude-opus-4.7),0.9846,0.8967,86
+3,2,1,LemonHarness_GPT-5.3-CodeX (),0.9231,0.8981,88
+4,9,-5,pilot-real (claude-opus-4-6),0.9143,0.8315,89
+5,5,0,Forge (GPT-5.4),0.8571,0.861,86
+6,7,-1,NexAU-AHE (gpt-5.5),0.8571,0.8517,89
+7,8,-1,CodeBrain-1.5 (GPT-5.3-Codex),0.8429,0.8494,87
+8,6,2,Capy (GPT-5.5),0.8286,0.8545,88
+9,4,5,JJAgent (Multiple),0.8135,0.8765,89
+10,10,0,OB-1_GPT-5.4-GPT-5.3-Codex-Claude-Opus-4.5-Claude-Opus-4.6 (),0.8,0.8247,89
+11,17,-6,Mux (GPT-5.3-Codex),0.8,0.7927,87
+12,16,-4,Terminus-KIRA (Gemini-3.1-Pro-Preview),0.7929,0.7973,86
+13,11,2,Polaris (Claude-Opus-4.7-GPT-5.5-Gemini-3.1-Pro),0.7857,0.8208,89
+14,21,-7,Deep-Agents (GPT-5.2-Codex),0.7857,0.7452,86
+15,31,-16,hookele (gpt5.1-codex-mini),0.7714,0.627,89
+16,19,-3,Junie_CLI (Gemini-3-Flash-Preview-Gemini-3.1-Pro-Preview-Claude-Opus-4.6-GPT-5.3-Codex),0.75,0.7545,86
+17,13,4,WOZCODE (Claude-Opus-4-7),0.7429,0.8169,89
+18,24,-6,OB-1_GPT-5.3-Codex-Claude-Opus-4.5-Claude-Opus-4.6 (),0.7429,0.7266,89
+19,25,-6,Crux (Claude-Opus-4.6),0.7202,0.7264,86
+20,12,8,Meta-Harness (Claude-Opus-4.6),0.7143,0.8194,85
+21,22,-1,Droid (Claude-Opus-4.6),0.7,0.731,87
+22,18,4,Codelia (GPT-5.3-Codex),0.6923,0.7891,87
+23,20,3,CodeBrain-1 (GPT-5.3-Codex),0.6769,0.7471,87
+24,15,9,LemonCode (GPT-5.3-Codex),0.6762,0.8045,89
+25,23,2,WozCode (Claude-Opus-4.6),0.6714,0.7272,87
+26,26,0,MAYA (Claude-4.6-opus),0.6714,0.7247,89
+27,28,-1,Ante (Gemini-3-Pro-Preview),0.6476,0.6755,87
+28,14,14,Droid (GPT-5.3-Codex),0.6462,0.8144,88
+29,27,2,clnkr (GPT-5.5),0.6429,0.7129,85
+30,30,0,CodeBrain-1 (Gemini-3-Pro-Preview),0.6167,0.6571,87
+31,32,-1,grok-cli (grok-4.20-0309-reasoning),0.6083,0.5812,86
+32,29,3,OpenSage (Gemini-3-Pro-Preview),0.5714,0.6614,88
+33,33,0,Gemini_CLI (Gemini-3-Flash-Preview),0.5154,0.5231,86
+34,35,-1,ClaudeCode (GLM-4.7),0.4615,0.3596,85
+35,34,1,MAYA (Claude-4.5-sonnet),0.4286,0.427,89
+36,36,0,little-coder (qwen3.6-35b-a3b),0.1857,0.2403,89
+37,37,0,little-coder (qwen3.5-9b),0.0643,0.091,89

mini_trials.csv

@@ -0,0 +1,509 @@
+agent,task,n_trials,pass_rate
+pilot-real (claude-opus-4-6),bn-fit-modify,5,1.0
+pilot-real (claude-opus-4-6),build-pmars,5,1.0
+pilot-real (claude-opus-4-6),crack-7z-hash,5,1.0
+pilot-real (claude-opus-4-6),custom-memory-heap-crash,5,1.0
+pilot-real (claude-opus-4-6),db-wal-recovery,5,1.0
+pilot-real (claude-opus-4-6),filter-js-from-html,5,1.0
+pilot-real (claude-opus-4-6),install-windows-3.11,5,0.8
+pilot-real (claude-opus-4-6),kv-store-grpc,5,1.0
+pilot-real (claude-opus-4-6),mcmc-sampling-stan,5,1.0
+pilot-real (claude-opus-4-6),mteb-retrieve,5,1.0
+pilot-real (claude-opus-4-6),qemu-alpine-ssh,5,0.0
+pilot-real (claude-opus-4-6),regex-log,5,1.0
+pilot-real (claude-opus-4-6),rstan-to-pystan,5,1.0
+pilot-real (claude-opus-4-6),vulnerable-secret,5,1.0
+little-coder (qwen3.5-9b),bn-fit-modify,5,0.0
+little-coder (qwen3.5-9b),build-pmars,5,0.2
+little-coder (qwen3.5-9b),crack-7z-hash,5,0.0
+little-coder (qwen3.5-9b),custom-memory-heap-crash,5,0.0
+little-coder (qwen3.5-9b),db-wal-recovery,5,0.0
+little-coder (qwen3.5-9b),filter-js-from-html,5,0.0
+little-coder (qwen3.5-9b),install-windows-3.11,5,0.0
+little-coder (qwen3.5-9b),kv-store-grpc,4,0.5
+little-coder (qwen3.5-9b),mcmc-sampling-stan,5,0.2
+little-coder (qwen3.5-9b),mteb-retrieve,5,0.0
+little-coder (qwen3.5-9b),qemu-alpine-ssh,4,0.0
+little-coder (qwen3.5-9b),regex-log,5,0.0
+little-coder (qwen3.5-9b),rstan-to-pystan,5,0.0
+little-coder (qwen3.5-9b),vulnerable-secret,5,0.0
+Meta-Harness (Claude-Opus-4.6),bn-fit-modify,5,1.0
+Meta-Harness (Claude-Opus-4.6),build-pmars,5,0.0
+Meta-Harness (Claude-Opus-4.6),crack-7z-hash,5,1.0
+Meta-Harness (Claude-Opus-4.6),custom-memory-heap-crash,5,1.0
+Meta-Harness (Claude-Opus-4.6),db-wal-recovery,4,1.0
+Meta-Harness (Claude-Opus-4.6),filter-js-from-html,5,0.0
+Meta-Harness (Claude-Opus-4.6),install-windows-3.11,5,0.2
+Meta-Harness (Claude-Opus-4.6),kv-store-grpc,5,1.0
+Meta-Harness (Claude-Opus-4.6),mcmc-sampling-stan,5,0.8
+Meta-Harness (Claude-Opus-4.6),mteb-retrieve,5,0.0
+Meta-Harness (Claude-Opus-4.6),qemu-alpine-ssh,5,1.0
+Meta-Harness (Claude-Opus-4.6),regex-log,5,1.0
+Meta-Harness (Claude-Opus-4.6),rstan-to-pystan,5,1.0
+Meta-Harness (Claude-Opus-4.6),vulnerable-secret,5,1.0
+grok-cli (grok-4.20-0309-reasoning),bn-fit-modify,1,1.0
+grok-cli (grok-4.20-0309-reasoning),build-pmars,6,1.0
+grok-cli (grok-4.20-0309-reasoning),crack-7z-hash,1,1.0
+grok-cli (grok-4.20-0309-reasoning),custom-memory-heap-crash,2,1.0
+grok-cli (grok-4.20-0309-reasoning),db-wal-recovery,2,0.5
+grok-cli (grok-4.20-0309-reasoning),filter-js-from-html,5,0.0
+grok-cli (grok-4.20-0309-reasoning),mcmc-sampling-stan,1,0.0
+grok-cli (grok-4.20-0309-reasoning),mteb-retrieve,5,0.0
+grok-cli (grok-4.20-0309-reasoning),qemu-alpine-ssh,5,0.8
+grok-cli (grok-4.20-0309-reasoning),regex-log,1,1.0
+grok-cli (grok-4.20-0309-reasoning),rstan-to-pystan,1,0.0
+grok-cli (grok-4.20-0309-reasoning),vulnerable-secret,6,1.0
+WozCode (Claude-Opus-4.6),bn-fit-modify,5,1.0
+WozCode (Claude-Opus-4.6),build-pmars,5,0.0
+WozCode (Claude-Opus-4.6),crack-7z-hash,5,1.0
+WozCode (Claude-Opus-4.6),custom-memory-heap-crash,5,1.0
+WozCode (Claude-Opus-4.6),db-wal-recovery,5,1.0
+WozCode (Claude-Opus-4.6),filter-js-from-html,5,0.0
+WozCode (Claude-Opus-4.6),install-windows-3.11,4,0.0
+WozCode (Claude-Opus-4.6),kv-store-grpc,5,1.0
+WozCode (Claude-Opus-4.6),mcmc-sampling-stan,5,1.0
+WozCode (Claude-Opus-4.6),mteb-retrieve,5,0.0
+WozCode (Claude-Opus-4.6),qemu-alpine-ssh,5,0.4
+WozCode (Claude-Opus-4.6),regex-log,5,1.0
+WozCode (Claude-Opus-4.6),rstan-to-pystan,5,1.0
+WozCode (Claude-Opus-4.6),vulnerable-secret,5,1.0
+WOZCODE (Claude-Opus-4-7),bn-fit-modify,5,1.0
+WOZCODE (Claude-Opus-4-7),build-pmars,5,1.0
+WOZCODE (Claude-Opus-4-7),crack-7z-hash,5,1.0
+WOZCODE (Claude-Opus-4-7),custom-memory-heap-crash,5,1.0
+WOZCODE (Claude-Opus-4-7),db-wal-recovery,5,1.0
+WOZCODE (Claude-Opus-4-7),filter-js-from-html,5,0.0
+WOZCODE (Claude-Opus-4-7),install-windows-3.11,4,0.0
+WOZCODE (Claude-Opus-4-7),kv-store-grpc,4,1.0
+WOZCODE (Claude-Opus-4-7),mcmc-sampling-stan,5,1.0
+WOZCODE (Claude-Opus-4-7),mteb-retrieve,5,0.4
+WOZCODE (Claude-Opus-4-7),qemu-alpine-ssh,2,1.0
+WOZCODE (Claude-Opus-4-7),regex-log,5,1.0
+WOZCODE (Claude-Opus-4-7),rstan-to-pystan,1,0.0
+WOZCODE (Claude-Opus-4-7),vulnerable-secret,5,1.0
+CodeBrain-1 (GPT-5.3-Codex),bn-fit-modify,5,1.0
+CodeBrain-1 (GPT-5.3-Codex),build-pmars,5,1.0
+CodeBrain-1 (GPT-5.3-Codex),crack-7z-hash,5,1.0
+CodeBrain-1 (GPT-5.3-Codex),custom-memory-heap-crash,5,1.0
+CodeBrain-1 (GPT-5.3-Codex),db-wal-recovery,5,0.2
+CodeBrain-1 (GPT-5.3-Codex),filter-js-from-html,4,0.0
+CodeBrain-1 (GPT-5.3-Codex),install-windows-3.11,5,0.0
+CodeBrain-1 (GPT-5.3-Codex),kv-store-grpc,5,0.4
+CodeBrain-1 (GPT-5.3-Codex),mcmc-sampling-stan,4,1.0
+CodeBrain-1 (GPT-5.3-Codex),mteb-retrieve,5,0.2
+CodeBrain-1 (GPT-5.3-Codex),regex-log,5,1.0
+CodeBrain-1 (GPT-5.3-Codex),rstan-to-pystan,3,1.0
+CodeBrain-1 (GPT-5.3-Codex),vulnerable-secret,5,1.0
+Droid (Claude-Opus-4.6),bn-fit-modify,5,1.0
+Droid (Claude-Opus-4.6),build-pmars,5,0.2
+Droid (Claude-Opus-4.6),crack-7z-hash,5,1.0
+Droid (Claude-Opus-4.6),custom-memory-heap-crash,5,1.0
+Droid (Claude-Opus-4.6),db-wal-recovery,1,0.0
+Droid (Claude-Opus-4.6),filter-js-from-html,5,0.0
+Droid (Claude-Opus-4.6),install-windows-3.11,5,0.8
+Droid (Claude-Opus-4.6),kv-store-grpc,5,1.0
+Droid (Claude-Opus-4.6),mcmc-sampling-stan,5,1.0
+Droid (Claude-Opus-4.6),mteb-retrieve,5,0.0
+Droid (Claude-Opus-4.6),qemu-alpine-ssh,5,0.8
+Droid (Claude-Opus-4.6),regex-log,5,1.0
+Droid (Claude-Opus-4.6),rstan-to-pystan,5,1.0
+Droid (Claude-Opus-4.6),vulnerable-secret,5,1.0
+LemonCode (GPT-5.3-Codex),bn-fit-modify,5,1.0
+LemonCode (GPT-5.3-Codex),build-pmars,5,1.0
+LemonCode (GPT-5.3-Codex),crack-7z-hash,5,0.8
+LemonCode (GPT-5.3-Codex),custom-memory-heap-crash,5,1.0
+LemonCode (GPT-5.3-Codex),db-wal-recovery,5,1.0
+LemonCode (GPT-5.3-Codex),filter-js-from-html,4,0.0
+LemonCode (GPT-5.3-Codex),install-windows-3.11,3,0.0
+LemonCode (GPT-5.3-Codex),kv-store-grpc,5,0.6
+LemonCode (GPT-5.3-Codex),mcmc-sampling-stan,5,0.4
+LemonCode (GPT-5.3-Codex),mteb-retrieve,2,1.0
+LemonCode (GPT-5.3-Codex),qemu-alpine-ssh,3,0.0
+LemonCode (GPT-5.3-Codex),regex-log,5,1.0
+LemonCode (GPT-5.3-Codex),rstan-to-pystan,3,0.6667
+LemonCode (GPT-5.3-Codex),vulnerable-secret,5,1.0
+Gemini_CLI (Gemini-3-Flash-Preview),bn-fit-modify,5,1.0
+Gemini_CLI (Gemini-3-Flash-Preview),build-pmars,4,0.5
+Gemini_CLI (Gemini-3-Flash-Preview),crack-7z-hash,4,1.0
+Gemini_CLI (Gemini-3-Flash-Preview),custom-memory-heap-crash,5,1.0
+Gemini_CLI (Gemini-3-Flash-Preview),db-wal-recovery,1,0.0
+Gemini_CLI (Gemini-3-Flash-Preview),filter-js-from-html,5,0.0
+Gemini_CLI (Gemini-3-Flash-Preview),install-windows-3.11,5,0.2
+Gemini_CLI (Gemini-3-Flash-Preview),kv-store-grpc,5,0.0
+Gemini_CLI (Gemini-3-Flash-Preview),mteb-retrieve,5,0.0
+Gemini_CLI (Gemini-3-Flash-Preview),qemu-alpine-ssh,2,0.0
+Gemini_CLI (Gemini-3-Flash-Preview),regex-log,5,1.0
+Gemini_CLI (Gemini-3-Flash-Preview),rstan-to-pystan,1,1.0
+Gemini_CLI (Gemini-3-Flash-Preview),vulnerable-secret,5,1.0
+vix (claude-opus-4-7),bn-fit-modify,5,1.0
+vix (claude-opus-4-7),build-pmars,5,1.0
+vix (claude-opus-4-7),crack-7z-hash,5,1.0
+vix (claude-opus-4-7),custom-memory-heap-crash,5,1.0
+vix (claude-opus-4-7),db-wal-recovery,5,1.0
+vix (claude-opus-4-7),filter-js-from-html,5,1.0
+vix (claude-opus-4-7),kv-store-grpc,5,1.0
+vix (claude-opus-4-7),mcmc-sampling-stan,5,1.0
+vix (claude-opus-4-7),mteb-retrieve,5,1.0
+vix (claude-opus-4-7),qemu-alpine-ssh,2,1.0
+vix (claude-opus-4-7),regex-log,5,1.0
+vix (claude-opus-4-7),rstan-to-pystan,5,1.0
+vix (claude-opus-4-7),vulnerable-secret,5,1.0
+Forge (GPT-5.4),bn-fit-modify,5,1.0
+Forge (GPT-5.4),build-pmars,5,1.0
+Forge (GPT-5.4),crack-7z-hash,5,1.0
+Forge (GPT-5.4),custom-memory-heap-crash,5,1.0
+Forge (GPT-5.4),db-wal-recovery,5,1.0
+Forge (GPT-5.4),filter-js-from-html,5,0.0
+Forge (GPT-5.4),install-windows-3.11,3,0.0
+Forge (GPT-5.4),kv-store-grpc,5,1.0
+Forge (GPT-5.4),mcmc-sampling-stan,5,1.0
+Forge (GPT-5.4),mteb-retrieve,5,1.0
+Forge (GPT-5.4),qemu-alpine-ssh,3,1.0
+Forge (GPT-5.4),regex-log,5,1.0
+Forge (GPT-5.4),rstan-to-pystan,4,1.0
+Forge (GPT-5.4),vulnerable-secret,5,1.0
+clnkr (GPT-5.5),bn-fit-modify,5,1.0
+clnkr (GPT-5.5),build-pmars,5,1.0
+clnkr (GPT-5.5),crack-7z-hash,5,1.0
+clnkr (GPT-5.5),custom-memory-heap-crash,3,0.0
+clnkr (GPT-5.5),db-wal-recovery,3,0.0
+clnkr (GPT-5.5),filter-js-from-html,4,0.0
+clnkr (GPT-5.5),install-windows-3.11,5,0.0
+clnkr (GPT-5.5),kv-store-grpc,5,1.0
+clnkr (GPT-5.5),mcmc-sampling-stan,5,1.0
+clnkr (GPT-5.5),mteb-retrieve,5,0.0
+clnkr (GPT-5.5),qemu-alpine-ssh,4,1.0
+clnkr (GPT-5.5),regex-log,5,1.0
+clnkr (GPT-5.5),rstan-to-pystan,5,1.0
+clnkr (GPT-5.5),vulnerable-secret,5,1.0
+Crux (Claude-Opus-4.6),bn-fit-modify,5,1.0
+Crux (Claude-Opus-4.6),build-pmars,5,0.6
+Crux (Claude-Opus-4.6),crack-7z-hash,5,1.0
+Crux (Claude-Opus-4.6),custom-memory-heap-crash,5,1.0
+Crux (Claude-Opus-4.6),db-wal-recovery,5,1.0
+Crux (Claude-Opus-4.6),filter-js-from-html,5,0.2
+Crux (Claude-Opus-4.6),install-windows-3.11,3,0.3333
+Crux (Claude-Opus-4.6),kv-store-grpc,5,1.0
+Crux (Claude-Opus-4.6),mcmc-sampling-stan,4,0.75
+Crux (Claude-Opus-4.6),mteb-retrieve,4,0.25
+Crux (Claude-Opus-4.6),qemu-alpine-ssh,5,0.8
+Crux (Claude-Opus-4.6),regex-log,5,0.4
+Crux (Claude-Opus-4.6),rstan-to-pystan,4,0.75
+Crux (Claude-Opus-4.6),vulnerable-secret,4,1.0
+ClaudeCode (GLM-4.7),bn-fit-modify,5,0.2
+ClaudeCode (GLM-4.7),build-pmars,5,0.4
+ClaudeCode (GLM-4.7),crack-7z-hash,5,1.0
+ClaudeCode (GLM-4.7),custom-memory-heap-crash,5,1.0
+ClaudeCode (GLM-4.7),db-wal-recovery,5,0.0
+ClaudeCode (GLM-4.7),filter-js-from-html,5,0.0
+ClaudeCode (GLM-4.7),install-windows-3.11,5,0.0
+ClaudeCode (GLM-4.7),kv-store-grpc,5,1.0
+ClaudeCode (GLM-4.7),mcmc-sampling-stan,4,1.0
+ClaudeCode (GLM-4.7),mteb-retrieve,5,0.0
+ClaudeCode (GLM-4.7),regex-log,5,0.2
+ClaudeCode (GLM-4.7),rstan-to-pystan,5,0.2
+ClaudeCode (GLM-4.7),vulnerable-secret,5,1.0
+NexAU-AHE (gpt-5.5),bn-fit-modify,5,1.0
+NexAU-AHE (gpt-5.5),build-pmars,5,1.0
+NexAU-AHE (gpt-5.5),crack-7z-hash,5,1.0
+NexAU-AHE (gpt-5.5),custom-memory-heap-crash,5,1.0
+NexAU-AHE (gpt-5.5),db-wal-recovery,5,1.0
+NexAU-AHE (gpt-5.5),filter-js-from-html,5,0.0
+NexAU-AHE (gpt-5.5),install-windows-3.11,5,0.4
+NexAU-AHE (gpt-5.5),kv-store-grpc,5,1.0
+NexAU-AHE (gpt-5.5),mcmc-sampling-stan,5,0.6
+NexAU-AHE (gpt-5.5),mteb-retrieve,5,1.0
+NexAU-AHE (gpt-5.5),qemu-alpine-ssh,4,1.0
+NexAU-AHE (gpt-5.5),regex-log,5,1.0
+NexAU-AHE (gpt-5.5),rstan-to-pystan,5,1.0
+NexAU-AHE (gpt-5.5),vulnerable-secret,5,1.0
+Deep-Agents (GPT-5.2-Codex),bn-fit-modify,5,0.8
+Deep-Agents (GPT-5.2-Codex),build-pmars,5,1.0
+Deep-Agents (GPT-5.2-Codex),crack-7z-hash,4,1.0
+Deep-Agents (GPT-5.2-Codex),custom-memory-heap-crash,5,1.0
+Deep-Agents (GPT-5.2-Codex),db-wal-recovery,3,1.0
+Deep-Agents (GPT-5.2-Codex),filter-js-from-html,5,0.0
+Deep-Agents (GPT-5.2-Codex),install-windows-3.11,5,0.0
+Deep-Agents (GPT-5.2-Codex),kv-store-grpc,5,1.0
+Deep-Agents (GPT-5.2-Codex),mcmc-sampling-stan,4,1.0
+Deep-Agents (GPT-5.2-Codex),mteb-retrieve,5,0.2
+Deep-Agents (GPT-5.2-Codex),qemu-alpine-ssh,2,1.0
+Deep-Agents (GPT-5.2-Codex),regex-log,5,1.0
+Deep-Agents (GPT-5.2-Codex),rstan-to-pystan,3,1.0
+Deep-Agents (GPT-5.2-Codex),vulnerable-secret,5,1.0
+logos-latest (claude-opus-4.7),bn-fit-modify,5,1.0
+logos-latest (claude-opus-4.7),build-pmars,5,0.8
+logos-latest (claude-opus-4.7),crack-7z-hash,5,1.0
+logos-latest (claude-opus-4.7),custom-memory-heap-crash,5,1.0
+logos-latest (claude-opus-4.7),db-wal-recovery,5,1.0
+logos-latest (claude-opus-4.7),filter-js-from-html,5,1.0
+logos-latest (claude-opus-4.7),kv-store-grpc,5,1.0
+logos-latest (claude-opus-4.7),mcmc-sampling-stan,5,1.0
+logos-latest (claude-opus-4.7),mteb-retrieve,5,1.0
+logos-latest (claude-opus-4.7),qemu-alpine-ssh,3,1.0
+logos-latest (claude-opus-4.7),regex-log,5,1.0
+logos-latest (claude-opus-4.7),rstan-to-pystan,5,1.0
+logos-latest (claude-opus-4.7),vulnerable-secret,5,1.0
+MAYA (Claude-4.5-sonnet),bn-fit-modify,5,0.0
+MAYA (Claude-4.5-sonnet),build-pmars,5,0.0
+MAYA (Claude-4.5-sonnet),crack-7z-hash,5,1.0
+MAYA (Claude-4.5-sonnet),custom-memory-heap-crash,5,1.0
+MAYA (Claude-4.5-sonnet),db-wal-recovery,5,1.0
+MAYA (Claude-4.5-sonnet),filter-js-from-html,5,1.0
+MAYA (Claude-4.5-sonnet),install-windows-3.11,5,0.0
+MAYA (Claude-4.5-sonnet),kv-store-grpc,5,0.0
+MAYA (Claude-4.5-sonnet),mcmc-sampling-stan,5,0.0
+MAYA (Claude-4.5-sonnet),mteb-retrieve,1,0.0
+MAYA (Claude-4.5-sonnet),qemu-alpine-ssh,5,0.0
+MAYA (Claude-4.5-sonnet),regex-log,5,1.0
+MAYA (Claude-4.5-sonnet),rstan-to-pystan,5,1.0
+MAYA (Claude-4.5-sonnet),vulnerable-secret,5,0.0
+CodeBrain-1 (Gemini-3-Pro-Preview),bn-fit-modify,5,1.0
+CodeBrain-1 (Gemini-3-Pro-Preview),build-pmars,3,0.3333
+CodeBrain-1 (Gemini-3-Pro-Preview),crack-7z-hash,3,1.0
+CodeBrain-1 (Gemini-3-Pro-Preview),custom-memory-heap-crash,5,1.0
+CodeBrain-1 (Gemini-3-Pro-Preview),db-wal-recovery,5,0.0
+CodeBrain-1 (Gemini-3-Pro-Preview),filter-js-from-html,5,0.0
+CodeBrain-1 (Gemini-3-Pro-Preview),install-windows-3.11,5,0.0
+CodeBrain-1 (Gemini-3-Pro-Preview),kv-store-grpc,5,1.0
+CodeBrain-1 (Gemini-3-Pro-Preview),mcmc-sampling-stan,5,1.0
+CodeBrain-1 (Gemini-3-Pro-Preview),mteb-retrieve,5,0.0
+CodeBrain-1 (Gemini-3-Pro-Preview),qemu-alpine-ssh,2,0.5
+CodeBrain-1 (Gemini-3-Pro-Preview),regex-log,5,1.0
+CodeBrain-1 (Gemini-3-Pro-Preview),rstan-to-pystan,5,0.8
+CodeBrain-1 (Gemini-3-Pro-Preview),vulnerable-secret,5,1.0
+Ante (Gemini-3-Pro-Preview),bn-fit-modify,5,0.8
+Ante (Gemini-3-Pro-Preview),build-pmars,5,0.4
+Ante (Gemini-3-Pro-Preview),crack-7z-hash,2,1.0
+Ante (Gemini-3-Pro-Preview),custom-memory-heap-crash,5,1.0
+Ante (Gemini-3-Pro-Preview),db-wal-recovery,5,0.0
+Ante (Gemini-3-Pro-Preview),filter-js-from-html,5,0.0
+Ante (Gemini-3-Pro-Preview),install-windows-3.11,5,0.0
+Ante (Gemini-3-Pro-Preview),kv-store-grpc,5,1.0
+Ante (Gemini-3-Pro-Preview),mcmc-sampling-stan,5,1.0
+Ante (Gemini-3-Pro-Preview),mteb-retrieve,5,0.6
+Ante (Gemini-3-Pro-Preview),qemu-alpine-ssh,3,0.6667
+Ante (Gemini-3-Pro-Preview),regex-log,5,1.0
+Ante (Gemini-3-Pro-Preview),rstan-to-pystan,5,0.6
+Ante (Gemini-3-Pro-Preview),vulnerable-secret,5,1.0
+Terminus-KIRA (Gemini-3.1-Pro-Preview),bn-fit-modify,5,1.0
+Terminus-KIRA (Gemini-3.1-Pro-Preview),build-pmars,5,1.0
+Terminus-KIRA (Gemini-3.1-Pro-Preview),crack-7z-hash,3,1.0
+Terminus-KIRA (Gemini-3.1-Pro-Preview),custom-memory-heap-crash,5,1.0
+Terminus-KIRA (Gemini-3.1-Pro-Preview),db-wal-recovery,2,0.5
+Terminus-KIRA (Gemini-3.1-Pro-Preview),filter-js-from-html,5,0.0
+Terminus-KIRA (Gemini-3.1-Pro-Preview),install-windows-3.11,5,0.0
+Terminus-KIRA (Gemini-3.1-Pro-Preview),kv-store-grpc,5,1.0
+Terminus-KIRA (Gemini-3.1-Pro-Preview),mcmc-sampling-stan,5,1.0
+Terminus-KIRA (Gemini-3.1-Pro-Preview),mteb-retrieve,5,1.0
+Terminus-KIRA (Gemini-3.1-Pro-Preview),qemu-alpine-ssh,4,1.0
+Terminus-KIRA (Gemini-3.1-Pro-Preview),regex-log,5,1.0
+Terminus-KIRA (Gemini-3.1-Pro-Preview),rstan-to-pystan,5,0.6
+Terminus-KIRA (Gemini-3.1-Pro-Preview),vulnerable-secret,5,1.0
+hookele (gpt5.1-codex-mini),bn-fit-modify,5,1.0
+hookele (gpt5.1-codex-mini),build-pmars,5,1.0
+hookele (gpt5.1-codex-mini),crack-7z-hash,5,1.0
+hookele (gpt5.1-codex-mini),custom-memory-heap-crash,5,1.0
+hookele (gpt5.1-codex-mini),db-wal-recovery,5,1.0
+hookele (gpt5.1-codex-mini),filter-js-from-html,5,0.0
+hookele (gpt5.1-codex-mini),install-windows-3.11,5,0.0
+hookele (gpt5.1-codex-mini),kv-store-grpc,5,1.0
+hookele (gpt5.1-codex-mini),mcmc-sampling-stan,5,0.8
+hookele (gpt5.1-codex-mini),mteb-retrieve,5,0.0
+hookele (gpt5.1-codex-mini),qemu-alpine-ssh,3,1.0
+hookele (gpt5.1-codex-mini),regex-log,5,1.0
+hookele (gpt5.1-codex-mini),rstan-to-pystan,3,1.0
+hookele (gpt5.1-codex-mini),vulnerable-secret,5,1.0
+LemonHarness_GPT-5.3-CodeX (),bn-fit-modify,5,1.0
+LemonHarness_GPT-5.3-CodeX (),build-pmars,5,1.0
+LemonHarness_GPT-5.3-CodeX (),crack-7z-hash,4,1.0
+LemonHarness_GPT-5.3-CodeX (),custom-memory-heap-crash,5,1.0
+LemonHarness_GPT-5.3-CodeX (),db-wal-recovery,5,1.0
+LemonHarness_GPT-5.3-CodeX (),filter-js-from-html,5,0.0
+LemonHarness_GPT-5.3-CodeX (),kv-store-grpc,5,1.0
+LemonHarness_GPT-5.3-CodeX (),mcmc-sampling-stan,4,1.0
+LemonHarness_GPT-5.3-CodeX (),mteb-retrieve,5,1.0
+LemonHarness_GPT-5.3-CodeX (),qemu-alpine-ssh,4,1.0
+LemonHarness_GPT-5.3-CodeX (),regex-log,5,1.0
+LemonHarness_GPT-5.3-CodeX (),rstan-to-pystan,4,1.0
+LemonHarness_GPT-5.3-CodeX (),vulnerable-secret,5,1.0
+OB-1_GPT-5.3-Codex-Claude-Opus-4.5-Claude-Opus-4.6 (),bn-fit-modify,5,1.0
+OB-1_GPT-5.3-Codex-Claude-Opus-4.5-Claude-Opus-4.6 (),build-pmars,5,1.0
+OB-1_GPT-5.3-Codex-Claude-Opus-4.5-Claude-Opus-4.6 (),crack-7z-hash,5,1.0
+OB-1_GPT-5.3-Codex-Claude-Opus-4.5-Claude-Opus-4.6 (),custom-memory-heap-crash,5,1.0
+OB-1_GPT-5.3-Codex-Claude-Opus-4.5-Claude-Opus-4.6 (),db-wal-recovery,5,0.0
+OB-1_GPT-5.3-Codex-Claude-Opus-4.5-Claude-Opus-4.6 (),filter-js-from-html,5,0.0
+OB-1_GPT-5.3-Codex-Claude-Opus-4.5-Claude-Opus-4.6 (),install-windows-3.11,5,0.0
+OB-1_GPT-5.3-Codex-Claude-Opus-4.5-Claude-Opus-4.6 (),kv-store-grpc,5,1.0
+OB-1_GPT-5.3-Codex-Claude-Opus-4.5-Claude-Opus-4.6 (),mcmc-sampling-stan,5,1.0
+OB-1_GPT-5.3-Codex-Claude-Opus-4.5-Claude-Opus-4.6 (),mteb-retrieve,5,0.6
+OB-1_GPT-5.3-Codex-Claude-Opus-4.5-Claude-Opus-4.6 (),qemu-alpine-ssh,5,1.0
+OB-1_GPT-5.3-Codex-Claude-Opus-4.5-Claude-Opus-4.6 (),regex-log,5,1.0
+OB-1_GPT-5.3-Codex-Claude-Opus-4.5-Claude-Opus-4.6 (),rstan-to-pystan,5,0.8
+OB-1_GPT-5.3-Codex-Claude-Opus-4.5-Claude-Opus-4.6 (),vulnerable-secret,5,1.0
+JJAgent (Multiple),bn-fit-modify,10,1.0
+JJAgent (Multiple),build-pmars,10,1.0
+JJAgent (Multiple),crack-7z-hash,10,1.0
+JJAgent (Multiple),custom-memory-heap-crash,10,1.0
+JJAgent (Multiple),db-wal-recovery,10,0.5
+JJAgent (Multiple),filter-js-from-html,10,0.0
+JJAgent (Multiple),install-windows-3.11,10,0.0
+JJAgent (Multiple),kv-store-grpc,10,1.0
+JJAgent (Multiple),mcmc-sampling-stan,10,1.0
+JJAgent (Multiple),mteb-retrieve,10,1.0
+JJAgent (Multiple),qemu-alpine-ssh,9,0.8889
+JJAgent (Multiple),regex-log,10,1.0
+JJAgent (Multiple),rstan-to-pystan,10,1.0
+JJAgent (Multiple),vulnerable-secret,10,1.0
+Codelia (GPT-5.3-Codex),bn-fit-modify,5,1.0
+Codelia (GPT-5.3-Codex),build-pmars,5,1.0
+Codelia (GPT-5.3-Codex),crack-7z-hash,5,1.0
+Codelia (GPT-5.3-Codex),custom-memory-heap-crash,5,1.0
+Codelia (GPT-5.3-Codex),db-wal-recovery,5,0.0
+Codelia (GPT-5.3-Codex),filter-js-from-html,5,0.0
+Codelia (GPT-5.3-Codex),install-windows-3.11,5,0.0
+Codelia (GPT-5.3-Codex),kv-store-grpc,5,0.8
+Codelia (GPT-5.3-Codex),mcmc-sampling-stan,5,1.0
+Codelia (GPT-5.3-Codex),mteb-retrieve,5,0.2
+Codelia (GPT-5.3-Codex),regex-log,5,1.0
+Codelia (GPT-5.3-Codex),rstan-to-pystan,5,1.0
+Codelia (GPT-5.3-Codex),vulnerable-secret,5,1.0
+Polaris (Claude-Opus-4.7-GPT-5.5-Gemini-3.1-Pro),bn-fit-modify,5,1.0
+Polaris (Claude-Opus-4.7-GPT-5.5-Gemini-3.1-Pro),build-pmars,5,1.0
+Polaris (Claude-Opus-4.7-GPT-5.5-Gemini-3.1-Pro),crack-7z-hash,5,1.0
+Polaris (Claude-Opus-4.7-GPT-5.5-Gemini-3.1-Pro),custom-memory-heap-crash,5,1.0
+Polaris (Claude-Opus-4.7-GPT-5.5-Gemini-3.1-Pro),db-wal-recovery,5,1.0
+Polaris (Claude-Opus-4.7-GPT-5.5-Gemini-3.1-Pro),filter-js-from-html,5,0.2
+Polaris (Claude-Opus-4.7-GPT-5.5-Gemini-3.1-Pro),install-windows-3.11,5,1.0
+Polaris (Claude-Opus-4.7-GPT-5.5-Gemini-3.1-Pro),kv-store-grpc,5,0.8
+Polaris (Claude-Opus-4.7-GPT-5.5-Gemini-3.1-Pro),mcmc-sampling-stan,5,0.6
+Polaris (Claude-Opus-4.7-GPT-5.5-Gemini-3.1-Pro),mteb-retrieve,5,0.4
+Polaris (Claude-Opus-4.7-GPT-5.5-Gemini-3.1-Pro),qemu-alpine-ssh,5,0.0
+Polaris (Claude-Opus-4.7-GPT-5.5-Gemini-3.1-Pro),regex-log,5,1.0
+Polaris (Claude-Opus-4.7-GPT-5.5-Gemini-3.1-Pro),rstan-to-pystan,5,1.0
+Polaris (Claude-Opus-4.7-GPT-5.5-Gemini-3.1-Pro),vulnerable-secret,5,1.0
+CodeBrain-1.5 (GPT-5.3-Codex),bn-fit-modify,5,0.8
+CodeBrain-1.5 (GPT-5.3-Codex),build-pmars,5,1.0
+CodeBrain-1.5 (GPT-5.3-Codex),crack-7z-hash,4,1.0
+CodeBrain-1.5 (GPT-5.3-Codex),custom-memory-heap-crash,5,1.0
+CodeBrain-1.5 (GPT-5.3-Codex),db-wal-recovery,5,1.0
+CodeBrain-1.5 (GPT-5.3-Codex),filter-js-from-html,4,0.0
+CodeBrain-1.5 (GPT-5.3-Codex),install-windows-3.11,5,0.0
+CodeBrain-1.5 (GPT-5.3-Codex),kv-store-grpc,5,1.0
+CodeBrain-1.5 (GPT-5.3-Codex),mcmc-sampling-stan,5,1.0
+CodeBrain-1.5 (GPT-5.3-Codex),mteb-retrieve,5,1.0
+CodeBrain-1.5 (GPT-5.3-Codex),qemu-alpine-ssh,3,1.0
+CodeBrain-1.5 (GPT-5.3-Codex),regex-log,5,1.0
+CodeBrain-1.5 (GPT-5.3-Codex),rstan-to-pystan,4,1.0
+CodeBrain-1.5 (GPT-5.3-Codex),vulnerable-secret,5,1.0
+Mux (GPT-5.3-Codex),bn-fit-modify,5,1.0
+Mux (GPT-5.3-Codex),build-pmars,5,1.0
+Mux (GPT-5.3-Codex),crack-7z-hash,5,1.0
+Mux (GPT-5.3-Codex),custom-memory-heap-crash,5,1.0
+Mux (GPT-5.3-Codex),db-wal-recovery,3,1.0
+Mux (GPT-5.3-Codex),filter-js-from-html,5,0.0
+Mux (GPT-5.3-Codex),install-windows-3.11,5,0.0
+Mux (GPT-5.3-Codex),kv-store-grpc,5,1.0
+Mux (GPT-5.3-Codex),mcmc-sampling-stan,5,1.0
+Mux (GPT-5.3-Codex),mteb-retrieve,5,0.2
+Mux (GPT-5.3-Codex),qemu-alpine-ssh,1,1.0
+Mux (GPT-5.3-Codex),regex-log,5,1.0
+Mux (GPT-5.3-Codex),rstan-to-pystan,5,1.0
+Mux (GPT-5.3-Codex),vulnerable-secret,5,1.0
+little-coder (qwen3.6-35b-a3b),bn-fit-modify,5,0.0
+little-coder (qwen3.6-35b-a3b),build-pmars,5,0.6
+little-coder (qwen3.6-35b-a3b),crack-7z-hash,5,0.8
+little-coder (qwen3.6-35b-a3b),custom-memory-heap-crash,5,0.0
+little-coder (qwen3.6-35b-a3b),db-wal-recovery,5,0.0
+little-coder (qwen3.6-35b-a3b),filter-js-from-html,4,0.0
+little-coder (qwen3.6-35b-a3b),install-windows-3.11,5,0.0
+little-coder (qwen3.6-35b-a3b),kv-store-grpc,5,0.0
+little-coder (qwen3.6-35b-a3b),mcmc-sampling-stan,5,0.2
+little-coder (qwen3.6-35b-a3b),mteb-retrieve,5,0.0
+little-coder (qwen3.6-35b-a3b),qemu-alpine-ssh,3,0.0
+little-coder (qwen3.6-35b-a3b),regex-log,5,0.0
+little-coder (qwen3.6-35b-a3b),rstan-to-pystan,5,0.0
+little-coder (qwen3.6-35b-a3b),vulnerable-secret,5,1.0
+OpenSage (Gemini-3-Pro-Preview),bn-fit-modify,5,1.0
+OpenSage (Gemini-3-Pro-Preview),build-pmars,4,0.0
+OpenSage (Gemini-3-Pro-Preview),crack-7z-hash,5,1.0
+OpenSage (Gemini-3-Pro-Preview),custom-memory-heap-crash,5,1.0
+OpenSage (Gemini-3-Pro-Preview),db-wal-recovery,5,0.0
+OpenSage (Gemini-3-Pro-Preview),filter-js-from-html,1,0.0
+OpenSage (Gemini-3-Pro-Preview),install-windows-3.11,5,0.0
+OpenSage (Gemini-3-Pro-Preview),kv-store-grpc,5,1.0
+OpenSage (Gemini-3-Pro-Preview),mcmc-sampling-stan,5,1.0
+OpenSage (Gemini-3-Pro-Preview),mteb-retrieve,5,0.0
+OpenSage (Gemini-3-Pro-Preview),qemu-alpine-ssh,4,0.0
+OpenSage (Gemini-3-Pro-Preview),regex-log,5,1.0
+OpenSage (Gemini-3-Pro-Preview),rstan-to-pystan,4,1.0
+OpenSage (Gemini-3-Pro-Preview),vulnerable-secret,5,1.0
+MAYA (Claude-4.6-opus),bn-fit-modify,5,1.0
+MAYA (Claude-4.6-opus),build-pmars,5,1.0
+MAYA (Claude-4.6-opus),crack-7z-hash,5,0.0
+MAYA (Claude-4.6-opus),custom-memory-heap-crash,5,1.0
+MAYA (Claude-4.6-opus),db-wal-recovery,5,0.0
+MAYA (Claude-4.6-opus),filter-js-from-html,3,0.0
+MAYA (Claude-4.6-opus),install-windows-3.11,5,0.0
+MAYA (Claude-4.6-opus),kv-store-grpc,5,1.0
+MAYA (Claude-4.6-opus),mcmc-sampling-stan,5,1.0
+MAYA (Claude-4.6-opus),mteb-retrieve,5,0.6
+MAYA (Claude-4.6-opus),qemu-alpine-ssh,5,1.0
+MAYA (Claude-4.6-opus),regex-log,5,1.0
+MAYA (Claude-4.6-opus),rstan-to-pystan,5,0.8
+MAYA (Claude-4.6-opus),vulnerable-secret,5,1.0
+Capy (GPT-5.5),bn-fit-modify,5,1.0
+Capy (GPT-5.5),build-pmars,5,1.0
+Capy (GPT-5.5),crack-7z-hash,5,1.0
+Capy (GPT-5.5),custom-memory-heap-crash,5,1.0
+Capy (GPT-5.5),db-wal-recovery,5,0.4
+Capy (GPT-5.5),filter-js-from-html,5,0.0
+Capy (GPT-5.5),install-windows-3.11,5,0.2
+Capy (GPT-5.5),kv-store-grpc,5,1.0
+Capy (GPT-5.5),mcmc-sampling-stan,4,1.0
+Capy (GPT-5.5),mteb-retrieve,5,1.0
+Capy (GPT-5.5),qemu-alpine-ssh,5,1.0
+Capy (GPT-5.5),regex-log,5,1.0
+Capy (GPT-5.5),rstan-to-pystan,5,1.0
+Capy (GPT-5.5),vulnerable-secret,5,1.0
+OB-1_GPT-5.4-GPT-5.3-Codex-Claude-Opus-4.5-Claude-Opus-4.6 (),bn-fit-modify,5,1.0
+OB-1_GPT-5.4-GPT-5.3-Codex-Claude-Opus-4.5-Claude-Opus-4.6 (),build-pmars,5,1.0
+OB-1_GPT-5.4-GPT-5.3-Codex-Claude-Opus-4.5-Claude-Opus-4.6 (),crack-7z-hash,5,1.0
+OB-1_GPT-5.4-GPT-5.3-Codex-Claude-Opus-4.5-Claude-Opus-4.6 (),custom-memory-heap-crash,5,1.0
+OB-1_GPT-5.4-GPT-5.3-Codex-Claude-Opus-4.5-Claude-Opus-4.6 (),db-wal-recovery,5,1.0
+OB-1_GPT-5.4-GPT-5.3-Codex-Claude-Opus-4.5-Claude-Opus-4.6 (),filter-js-from-html,5,0.0
+OB-1_GPT-5.4-GPT-5.3-Codex-Claude-Opus-4.5-Claude-Opus-4.6 (),install-windows-3.11,5,0.0
+OB-1_GPT-5.4-GPT-5.3-Codex-Claude-Opus-4.5-Claude-Opus-4.6 (),kv-store-grpc,5,1.0
+OB-1_GPT-5.4-GPT-5.3-Codex-Claude-Opus-4.5-Claude-Opus-4.6 (),mcmc-sampling-stan,5,1.0
+OB-1_GPT-5.4-GPT-5.3-Codex-Claude-Opus-4.5-Claude-Opus-4.6 (),mteb-retrieve,5,0.2
+OB-1_GPT-5.4-GPT-5.3-Codex-Claude-Opus-4.5-Claude-Opus-4.6 (),qemu-alpine-ssh,5,1.0
+OB-1_GPT-5.4-GPT-5.3-Codex-Claude-Opus-4.5-Claude-Opus-4.6 (),regex-log,5,1.0
+OB-1_GPT-5.4-GPT-5.3-Codex-Claude-Opus-4.5-Claude-Opus-4.6 (),rstan-to-pystan,5,1.0
+OB-1_GPT-5.4-GPT-5.3-Codex-Claude-Opus-4.5-Claude-Opus-4.6 (),vulnerable-secret,5,1.0
+Droid (GPT-5.3-Codex),bn-fit-modify,5,1.0
+Droid (GPT-5.3-Codex),build-pmars,5,1.0
+Droid (GPT-5.3-Codex),crack-7z-hash,5,1.0
+Droid (GPT-5.3-Codex),custom-memory-heap-crash,5,1.0
+Droid (GPT-5.3-Codex),db-wal-recovery,5,0.4
+Droid (GPT-5.3-Codex),filter-js-from-html,5,0.0
+Droid (GPT-5.3-Codex),install-windows-3.11,5,0.0
+Droid (GPT-5.3-Codex),kv-store-grpc,5,1.0
+Droid (GPT-5.3-Codex),mcmc-sampling-stan,5,0.0
+Droid (GPT-5.3-Codex),mteb-retrieve,5,0.0
+Droid (GPT-5.3-Codex),regex-log,5,1.0
+Droid (GPT-5.3-Codex),rstan-to-pystan,5,1.0
+Droid (GPT-5.3-Codex),vulnerable-secret,5,1.0
+Junie_CLI (Gemini-3-Flash-Preview-Gemini-3.1-Pro-Preview-Claude-Opus-4.6-GPT-5.3-Codex),bn-fit-modify,5,0.8
+Junie_CLI (Gemini-3-Flash-Preview-Gemini-3.1-Pro-Preview-Claude-Opus-4.6-GPT-5.3-Codex),build-pmars,5,0.8
+Junie_CLI (Gemini-3-Flash-Preview-Gemini-3.1-Pro-Preview-Claude-Opus-4.6-GPT-5.3-Codex),crack-7z-hash,4,1.0
+Junie_CLI (Gemini-3-Flash-Preview-Gemini-3.1-Pro-Preview-Claude-Opus-4.6-GPT-5.3-Codex),custom-memory-heap-crash,5,1.0
+Junie_CLI (Gemini-3-Flash-Preview-Gemini-3.1-Pro-Preview-Claude-Opus-4.6-GPT-5.3-Codex),db-wal-recovery,2,0.5
+Junie_CLI (Gemini-3-Flash-Preview-Gemini-3.1-Pro-Preview-Claude-Opus-4.6-GPT-5.3-Codex),filter-js-from-html,5,0.0
+Junie_CLI (Gemini-3-Flash-Preview-Gemini-3.1-Pro-Preview-Claude-Opus-4.6-GPT-5.3-Codex),install-windows-3.11,5,0.0
+Junie_CLI (Gemini-3-Flash-Preview-Gemini-3.1-Pro-Preview-Claude-Opus-4.6-GPT-5.3-Codex),kv-store-grpc,5,1.0
+Junie_CLI (Gemini-3-Flash-Preview-Gemini-3.1-Pro-Preview-Claude-Opus-4.6-GPT-5.3-Codex),mcmc-sampling-stan,4,1.0
+Junie_CLI (Gemini-3-Flash-Preview-Gemini-3.1-Pro-Preview-Claude-Opus-4.6-GPT-5.3-Codex),mteb-retrieve,5,0.4
+Junie_CLI (Gemini-3-Flash-Preview-Gemini-3.1-Pro-Preview-Claude-Opus-4.6-GPT-5.3-Codex),qemu-alpine-ssh,3,1.0
+Junie_CLI (Gemini-3-Flash-Preview-Gemini-3.1-Pro-Preview-Claude-Opus-4.6-GPT-5.3-Codex),regex-log,5,1.0
+Junie_CLI (Gemini-3-Flash-Preview-Gemini-3.1-Pro-Preview-Claude-Opus-4.6-GPT-5.3-Codex),rstan-to-pystan,5,1.0
+Junie_CLI (Gemini-3-Flash-Preview-Gemini-3.1-Pro-Preview-Claude-Opus-4.6-GPT-5.3-Codex),vulnerable-secret,5,1.0

size_curve.csv

@@ -0,0 +1,11 @@
+n_tasks,mean_agreement,sd,p10,p90,best,worst,n_samples
+4,0.7816,0.0401,0.723,0.8348,0.8604,0.6629,200
+6,0.8009,0.0429,0.7432,0.8521,0.8851,0.6246,200
+8,0.8126,0.0418,0.756,0.8596,0.8956,0.6824,200
+10,0.8283,0.0402,0.7703,0.8724,0.9099,0.7042,200
+14,0.8536,0.0347,0.8063,0.8941,0.9354,0.7327,200
+18,0.8695,0.0296,0.8296,0.9032,0.9287,0.7515,200
+22,0.8852,0.0249,0.8514,0.9129,0.9347,0.7958,200
+30,0.907,0.0192,0.8859,0.9302,0.9489,0.8408,200
+40,0.925,0.0141,0.9077,0.9422,0.9662,0.8844,200
+60,0.9506,0.0101,0.9384,0.964,0.9775,0.9152,200

tasks.csv

@@ -0,0 +1,15 @@
+task,field_pass_rate,spread_across_agents,trial_noise,n_agents
+filter-js-from-html,0.1189,0.31,0.0222,37
+install-windows-3.11,0.1192,0.2583,0.0897,33
+mteb-retrieve,0.4122,0.4203,0.1492,37
+db-wal-recovery,0.5676,0.4485,0.0966,37
+qemu-alpine-ssh,0.6926,0.4169,0.093,33
+build-pmars,0.7523,0.3611,0.1225,37
+mcmc-sampling-stan,0.8153,0.328,0.1001,36
+rstan-to-pystan,0.8167,0.3292,0.1025,37
+kv-store-grpc,0.8639,0.2974,0.0633,36
+bn-fit-modify,0.8757,0.2954,0.0556,37
+regex-log,0.9081,0.2695,0.0247,37
+custom-memory-heap-crash,0.9189,0.273,0.0,37
+crack-7z-hash,0.9351,0.2281,0.0222,37
+vulnerable-secret,0.9459,0.2261,0.0,37