✨ Add openclaw benchmarks

app.py

@@ -49,6 +49,7 @@ from huggingface_hub import HfApi
 
 from src.leaderboard import get_leaderboard_df, get_benchmark_run_df
 from src.display.text_blocks import (
+    HOW_TO_USE_TEXT,
     INTRODUCTION_TEXT,
     LLM_BENCHMARKS_TEXT,
 )
@@ -175,6 +176,8 @@ with demo:
 
         with gr.Tab("📝 About"):
             gr.Markdown(LLM_BENCHMARKS_TEXT, elem_classes="markdown-text")
+    
+    gr.Markdown(HOW_TO_USE_TEXT, elem_classes="markdown-text")
 
 scheduler = BackgroundScheduler()
 scheduler.add_job(restart_space, "interval", seconds=1800)

results/swe-bench-pro--ansible-qwen3-6-36b-nvfp4-openclaw.json

@@ -0,0 +1,60 @@
+{
+    "benchmark": {
+        "name": "swe-bench-pro--ansible",
+        "repo": "ScaleAI/SWE-bench_Pro",
+        "num_tasks": 96,
+        "url": "https://huggingface.co/datasets/ScaleAI/SWE-bench_Pro"
+    },
+    "harness": {
+        "name": "OpenClaw",
+        "skills": [],
+        "is_oss": true,
+        "url": "https://github.com/openclaw/openclaw"
+    },
+    "model": {
+        "name": "Qwen3.6-35B-A3B",
+        "repo": "RedHatAI/Qwen3.6-35B-A3B-NVFP4",
+        "is_oss": true,
+        "num_params": 35,
+        "precision": "nvfp4",
+        "url": "https://huggingface.co/RedHatAI/Qwen3.6-35B-A3B-NVFP4"
+    },
+    "environment": {
+        "name": "harbor",
+        "config": {
+            "path": null,
+            "name": "scale-ai/swe-bench-pro",
+            "version": null,
+            "ref": "sha256:88411d32ff27e53a4c1a7e29f0c2aeba180c8e5d60f221cab5ed56325f33549d",
+            "registry_url": null,
+            "registry_path": null,
+            "overwrite": false,
+            "download_dir": null,
+            "task_names": [
+                "*ansible*"
+            ],
+            "exclude_task_names": null,
+            "n_tasks": null
+        },
+        "url": "https://github.com/harbor-framework/harbor"
+    },
+    "metrics": {
+        "n_tasks": 96,
+        "n_errors": 5,
+        "score": 0.406,
+        "n_input_tokens": 0,
+        "n_cache_tokens": 0,
+        "n_output_tokens": 0,
+        "n_total_tokens": 0,
+        "agent_time_seconds": 38085,
+        "total_time_seconds": 50779,
+        "cost_usd": 9.4,
+        "mean_input_tokens_per_task": 0,
+        "mean_cache_tokens_per_task": 0,
+        "mean_output_tokens_per_task": 0,
+        "mean_tokens_per_task": 0,
+        "mean_cost_usd_per_task": 0.1,
+        "mean_total_time_seconds_per_task": 528,
+        "mean_agent_time_seconds_per_task": 396
+    }
+}

results/swe-bench-verified-qwen3-6-35b-nvfp4-openclaw.json

@@ -0,0 +1,59 @@
+{
+    "benchmark": {
+        "name": "swe-bench-verified",
+        "repo": "SWE-bench/SWE-bench_Verified",
+        "num_tasks": 500,
+        "url": "https://huggingface.co/datasets/SWE-bench/SWE-bench_Verified"
+    },
+    "harness": {
+        "name": "OpenClaw",
+        "skills": [],
+        "is_oss": true,
+        "url": "https://github.com/openclaw/openclaw"
+    },
+    "model": {
+        "name": "Qwen3.6-35B-A3B",
+        "repo": "RedHatAI/Qwen3.6-35B-A3B-NVFP4",
+        "is_oss": true,
+        "num_params": 35,
+        "precision": "nvfp4",
+        "url": "https://huggingface.co/RedHatAI/Qwen3.6-35B-A3B-NVFP4"
+    },
+    "environment": {
+        "name": "harbor",
+        "config": {
+            "path": null,
+            "name": "swe-bench/swe-bench-verified",
+            "version": null,
+            "ref": "sha256:235d6032d549851a936db3b5fe08807c4d385c12ee10e7be9c9786a1ff60563c",
+            "registry_url": null,
+            "registry_path": null,
+            "overwrite": false,
+            "download_dir": null,
+            "task_names": null,
+            "exclude_task_names": null,
+            "n_tasks": null,
+            "accelerated_images": true
+        },
+        "url": "https://github.com/harbor-framework/harbor"
+    },
+    "metrics": {
+        "n_tasks": 500,
+        "n_errors": 3,
+        "score": 0.588,
+        "n_input_tokens": 0,
+        "n_cache_tokens": 0,
+        "n_output_tokens": 0,
+        "n_total_tokens": 0,
+        "agent_time_seconds": 120399,
+        "total_time_seconds": 200354,
+        "cost_usd": 33.44,
+        "mean_input_tokens_per_task": 0,
+        "mean_cache_tokens_per_task": 0,
+        "mean_output_tokens_per_task": 0,
+        "mean_tokens_per_task": 0,
+        "mean_cost_usd_per_task": 0.07,
+        "mean_total_time_seconds_per_task": 400,
+        "mean_agent_time_seconds_per_task": 240
+    }
+}

src/display/text_blocks.py

@@ -32,3 +32,22 @@ A coding agent is a system that autonomously solves software engineering tasks -
 
 Visit the [GitHub repo](https://github.com/redhat-et/coding_agent_bench) for details about the project, methodology, and how to submit your own results.
 """
+
+HOW_TO_USE_TEXT = """
+---
+## How to interpret these results
+
+In the absence of enterprise-specific datasets, public benchmarks provide a means of comparing the performance of coding agents across a wide range of tasks.
+Better performance on these benchmarks generally translates to better performance on real-world tasks.
+All benchmarks are run using Harbor, a sandboxed environment for evaluating coding agents.
+
+Each benchmark measures the performance of the coding agent on different tasks:
+
+* **swe-bench-verified**: Measures performance on solving GitHub issues in popular Python repositories.
+* **swe-bench-pro--ansible**: Measures performance on solving GitHub issues in the [ansible/ansible](https://github.com/ansible/ansible) repository. 
+    Demonstrates how benchmarking can be used to evaluate coding agents on enterprise-specific tasks.
+
+Higher scores indicate better performance on the benchmarks.
+If an agent scores better on a given benchmark than another, it can be generally considered to be better at those kinds of tasks.
+We take a simple average of these scores so you can quickly compare the performance of different coding agents, but this is a relative score and the average itself is meaningless on its own.
+"""