Spaces:

taagarwa
/

coding-agent-leaderboard

Running

App Files Files Community

taagarwa commited on 16 days ago

Commit

7a6725b

1 Parent(s): 456b71c

✨ Add visual formatting and themeing

Browse files

Files changed (2) hide show

app.py +31 -10
src/display/text_blocks.py +27 -5

app.py CHANGED Viewed

@@ -18,14 +18,6 @@ REPO_ID = "taagarwa/coding-agent-leaderboard"
 TOKEN = os.environ.get("HF_TOKEN")
 API = HfApi(token=TOKEN)
-HEAD = """
-<head>
-    <base target="_blank">
-</head>
-<h1>Coding Agent Leaderboard</h1>
-"""
 def restart_space():
     API.restart_space(repo_id=REPO_ID)
@@ -36,6 +28,35 @@ def extract_body(s: str):
     return re.match(r'\[(.*?)\]', s).group(1)
 def init_leaderboard(dataframe):
     if dataframe is None or dataframe.empty:
         raise ValueError("Leaderboard DataFrame is empty or None.")
@@ -61,9 +82,9 @@ def init_leaderboard(dataframe):
         interactive=False,
     )
-demo = gr.Blocks()
 with demo:
-    gr.HTML(HEAD)
     gr.Markdown(INTRODUCTION_TEXT, elem_classes="markdown-text")
     with gr.Tabs() as tabs:

 TOKEN = os.environ.get("HF_TOKEN")
 API = HfApi(token=TOKEN)
 def restart_space():
     API.restart_space(repo_id=REPO_ID)
     return re.match(r'\[(.*?)\]', s).group(1)
+def build_header_html(df):
+    n_results = len(df)
+    n_models = df["model"].nunique()
+    n_harnesses = df["harness"].apply(lambda s: extract_body(s)).nunique()
+    n_datasets = df["dataset"].apply(lambda s: extract_body(s)).nunique()
+    return f"""
+    <base target="_blank">
+    <div style="padding: 1.5rem 0.5rem 1rem 0.5rem; text-align: left;">
+        <h1 style="margin: 0 0 0.5rem 0; font-size: 2rem;">
+            Coding Agent Leaderboard
+        </h1>
+        <div style="height: 4px; border-radius: 2px; background: linear-gradient(90deg, #84cc16, #f59e0b); margin-bottom: 0.75rem;"></div>
+        <p style="margin: 0 0 0.75rem 0; font-size: 1.1rem; opacity: 0.8;">
+            Compare coding agents across models, harnesses, and environments
+        </p>
+        <div style="display: flex; gap: 0.5rem; flex-wrap: wrap; font-size: 0.95rem; opacity: 0.7;">
+            <span style="font-weight: 600;">{n_results} Results</span>
+            <span>·</span>
+            <span style="font-weight: 600;">{n_models} Models</span>
+            <span>·</span>
+            <span style="font-weight: 600;">{n_harnesses} Harnesses</span>
+            <span>·</span>
+            <span style="font-weight: 600;">{n_datasets} Datasets</span>
+        </div>
+    </div>
+    """
 def init_leaderboard(dataframe):
     if dataframe is None or dataframe.empty:
         raise ValueError("Leaderboard DataFrame is empty or None.")
         interactive=False,
     )
+demo = gr.Blocks(theme="citrus")
 with demo:
+    gr.HTML(build_header_html(LEADERBOARD_DF))
     gr.Markdown(INTRODUCTION_TEXT, elem_classes="markdown-text")
     with gr.Tabs() as tabs:

src/display/text_blocks.py CHANGED Viewed

@@ -1,15 +1,37 @@
 INTRODUCTION_TEXT = """
-Welcome to the Coding Agent Leaderboard!
 """
 LLM_BENCHMARKS_TEXT = """
-## About
-Evaluate and compare Coding Agents.
-Coding Agent = Model + Harness + Skills.
-Visit our [GitHub repo](https://github.com/redhat-et/coding_agent_bench) for more details about the project.
 """
 CITATION_BUTTON_TEXT = "TBD"

 INTRODUCTION_TEXT = """
+A **Coding Agent** is more than just a model - it's the combination of a **Model**, a **Harness** (the tool/framework driving the model), and **Skills** (the instructions that guide the agent's behavior).
+This leaderboard tracks how these components work together, because the same model can perform very differently depending on the harness and skills it's paired with.
 """
 LLM_BENCHMARKS_TEXT = """
+## What is a Coding Agent?
+A coding agent is a system that autonomously solves software engineering tasks - reading code, reasoning about bugs, and writing patches. Its performance depends on three components:
+- **Model** - The underlying language model (e.g. Claude Opus 4.7, Qwen3.6-35B)
+- **Harness** - The framework or tool that orchestrates the model's actions (e.g. Claude Code, OpenCode, Pi)
+- **Skills** - The instructions guiding the agent's behavior
+## How to Read the Table
+| Column | Description |
+|--------|-------------|
+| **Dataset** | The benchmark used for evaluation (e.g. SWE-bench Verified - 500 real GitHub issues) |
+| **Harness** | The agent framework driving the model. Entries marked with `*` are **internal** - the provider ran the benchmark but did not publish the harness or environment |
+| **Model** | The language model being evaluated |
+| **Environment** | The benchmark runtime. Also marked `*` when internal |
+| **Score** | Outcome of the benchmark, often the fraction of tasks solved correctly (higher is better) |
+| **Precision** | Model weight format (e.g. bf16, fp4) - affects speed, memory footprint, and quality |
+## Key Concepts
+- **FOSS vs Proprietary** - Filters let you compare fully open-source agents against proprietary ones. A FOSS model with a FOSS harness means anyone can reproduce the result
+- **Skills** - Some harnesses augment the model with extra capabilities (tools, retrieval, etc.). Listed in the "skills" column when present
+- **Internal results (`*`)** - Benchmarks run by the model provider where the harness and environment were not made public. These are useful reference points but are not independently reproducible
+## Learn More
+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.
 """
 CITATION_BUTTON_TEXT = "TBD"