Upload folder using huggingface_hub

README.md

@@ -0,0 +1,52 @@
+---
+license: mit
+tags:
+- agent-benchmark
+- code-agents
+- llm-evaluation
+- openhands
+- goose
+- harbor
+---
+
+# Harbor Goose vs OpenHands Benchmark
+
+Trial-level results from a small controlled study comparing two agent harnesses —
+[Goose](https://github.com/block/goose) and [OpenHands-SDK](https://github.com/All-Hands-AI/OpenHands) —
+on a frozen 40-task [Harbor](https://github.com/harbor-framework/harbor) Terminal-Bench-Pro slice.
+
+All runs used `minimax/minimax-m2.5` via OpenRouter with Daytona as the sandbox backend.
+
+## Key Findings
+
+- Cutting turns 100→60: Goose improved (0.45→0.525), OpenHands dropped (0.575→0.500) — **opposite directions**
+- The tweaked prompt fully recovered OpenHands' loss at 60 turns; had no effect on Goose
+- At their best both harnesses reach **0.575** pass rate
+- OpenHands used **35× more tokens** per trial (~500K vs ~14K) to reach the same score
+
+## Files
+
+| File | Description |
+|---|---|
+| `trials.csv` | One row per trial (240 rows across 6 variants) |
+| `trials.jsonl` | Same data in JSONL format |
+| `variant_summary.csv` | Aggregated pass rate, tokens, timing per variant |
+
+## Variants
+
+| Variant | Harness | Turns | Prompt | Pass Rate |
+|---|---|---|---|---|
+| goose-base | Goose | 100 | base | 0.450 |
+| goose-base-60 | Goose | 60 | base | 0.525 |
+| goose-tweaked | Goose | 60 | tweaked | 0.500 |
+| openhands-sdk-base | OpenHands-SDK | 100 | base | 0.575 |
+| openhands-sdk-base-60 | OpenHands-SDK | 60 | base | 0.500 |
+| openhands-sdk-tweaked | OpenHands-SDK | 60 | tweaked | 0.575 |
+
+## Caveats
+
+n=40 tasks. A 2–3 task swing is within run-to-run variance. No statistical significance testing performed.
+
+## Full Code & Configs
+
+GitHub: https://github.com/namanvats/harness-experiments-harbor

variant_summary.csv

@@ -0,0 +1,7 @@
+harness_way,trials,correct,failed,pass_rate,avg_reward,turn_limit_values,avg_turn_limit,sum_input_tokens,sum_output_tokens,sum_total_tokens,rows_with_total_tokens,avg_input_tokens,avg_output_tokens,avg_total_tokens,estimated_token_rows,avg_trial_seconds,avg_agent_execution_seconds
+goose-base,40,18,22,0.45,0.461538,100,100.0,452163,193783,645946,35,12918.94,5536.66,18455.6,35,498.87,461.35
+goose-base-60,40,21,19,0.525,0.538462,60,60.0,367987,157710,525697,37,9945.59,4262.43,14208.03,37,362.4,299.77
+goose-tweaked,40,20,20,0.5,0.512821,60,60.0,360667,154574,515241,37,9747.76,4177.68,13925.43,37,349.21,279.95
+openhands-sdk-base,40,23,17,0.575,0.605263,100,100.0,16620443,270059,16890502,33,503649.79,8183.61,511833.39,0,501.03,439.2
+openhands-sdk-base-60,40,20,20,0.5,0.526316,60,60.0,17756494,287932,18044426,36,493235.94,7998.11,501234.06,0,389.76,325.05
+openhands-sdk-tweaked,40,23,17,0.575,0.605263,60,60.0,20845914,358187,21204101,37,563403.08,9680.73,573083.81,0,336.42,278.16