Leaderboard data: counted attempts, numbers and links to every run

README.md

@@ -12,6 +12,8 @@ tags:
 
 This repository holds the pictures, videos and agent runs behind the [Embodied First Exam](https://www.embodied-first-exam.ai), a benchmark for code agents in robotics. Each question in the exam is a frozen simulation scene with a goal written for a person, and the answer is a program, which is marked by replaying it in a fresh simulator.
 
+Leaderboard data (every counted attempt of the ranked agents, with a link to its full run): [`leaderboard/`](leaderboard/README.md).
+
 ## Layout
 
 ```

leaderboard/README.md

@@ -0,0 +1,72 @@
+# Embodied First Exam: leaderboard data
+
+The numbers behind the [leaderboard](https://www.embodied-first-exam.ai/leaderboard/), one row per counted attempt, and a
+link from every attempt to its full run. Snapshot of 5 November 2026; the newest counted attempt finished on
+2 November 2026. Task set: the exam repository [mll-lab-nu/embodied-first-exam](https://github.com/mll-lab-nu/embodied-first-exam)
+at main `ebd10b62b0`.
+
+## What is counted
+
+| | tasks | tests (task x mode) | privileged | standard | realtime | suites |
+|---|---|---|---|---|---|---|
+| all | 1,156 | 2,288 | 1,047 | 1,155 | 86 | 35 |
+| core | 456 | 901 | 391 | 456 | 54 | |
+
+- **Core** = `docs/core_set.toml` of that commit. Each core task has a default mode: 403 standard, 53 realtime.
+- **One attempt per (agent, test):** the latest finished valid attempt, success or failed. Infrastructure failures
+  caught before grading are re-run and do not count; ablation and quick-check runs are left out. An attempt that an
+  API or infrastructure error ended after grading keeps its verdict and is labelled in its run.
+- **Success** = verifier reward >= 1 within the token budget.
+- **Missing counts as failed:** a test with no counted attempt is a failure (59 of the 6,307 pairs).
+- **Views:** Core = each core task's default test (456). Privileged / Standard / Realtime = every core test of that
+  mode (391 / 456 / 54). Rate = solved / tests of the view.
+
+| agent | Core /456 | Privileged /391 | Standard /456 | Realtime /54 | missing |
+|---|---|---|---|---|---|
+| Codex · GPT-6 Astra | 150 (32.9%) | 303 (77.5%) | 156 (34.2%) | 17 (31.5%) | 3 |
+| Claude Code · Opus 5.5 | 107 (23.5%) | 231 (59.1%) | 114 (25.0%) | 4 (7.4%) | 1 |
+| Antigravity · Gemini 4 Argon | 89 (19.5%) | 164 (41.9%) | 96 (21.1%) | 4 (7.4%) | 0 |
+| Antigravity · Gemini 3.8 Flash | 27 (5.9%) | 96 (24.6%) | 27 (5.9%) | 1 (1.9%) | 0 |
+| Claude Code · Qwen3.8 Max | 18 (3.9%) | 37 (9.5%) | 18 (3.9%) | 1 (1.9%) | 48 |
+| dsh · DeepSeek V4.1 Flash | 12 (2.6%) | 69 (17.6%) | 12 (2.6%) | 1 (1.9%) | 0 |
+| Claude Code · GLM 5.3 Flash | 11 (2.4%) | 36 (9.2%) | 11 (2.4%) | 0 (0.0%) | 7 |
+
+"missing" counts tests over all three modes.
+
+## Files
+
+| file | content |
+|---|---|
+| `numbers.json` | the sizes above, the rule, and per agent and view: solved, tests, rate, missing |
+| `leaderboard_core.csv` | 901 core tests x 7 agents = 6,307 rows: verdict (`success` / `failed` / `missing`), `counts_as`, the attempt's `trial_id` and `hf_slug` |
+| `trials.csv` | the 6,248 counted attempts: suite, task dir, setting, is_default, run, verdict, reward, finished_at (UTC), `in_common7` (the 843 core tests all seven agents have a result on), `hf_slug`, `has_video`, `hf_run_dir`, `viewer_url`, `video_url` |
+| `tests.csv` | the 901 core tests: default or not, in the common set or not, and which runs lack a result |
+| `check_coverage.py` | checks that every row of `trials.csv` has its run in [MLL-Lab/embodied-exam-runs](https://huggingface.co/datasets/MLL-Lab/embodied-exam-runs) with the same outcome |
+
+`run_id` is `<harness>-<model>-<effort>`, e.g. `codex-gpt6_astra-high`. A task dir is `<suite>-<task>-<mode>`.
+
+## Opening a run
+
+Every counted attempt has a run in the public dataset `MLL-Lab/embodied-exam-runs`:
+
+- slug = `<task dir>--<last 7 characters of trial_id>`, lower case (the `hf_slug` column);
+- viewer: `https://www.embodied-first-exam.ai/run/?r=<slug>`;
+- files: `runs/<first 2 hex digits of sha1(slug)>/<slug>/`: `run.json` (every step, tool call and output),
+  `trajectory.jsonl.gz` (the agent's full trajectory), `steps.bin` (the images the agent saw), and `replay.mp4` +
+  `poster.jpg` when the attempt has a video (privileged attempts that submitted nothing have none).
+
+```
+huggingface-cli download MLL-Lab/embodied-exam-runs --repo-type dataset \
+  --include "runs/e1/ambench-cabinet-pick-place-fahexa-i00-privileged--zvvkvgc/*" --local-dir efe-runs
+python3 check_coverage.py trials.csv      # no token needed; HF may rate-limit anonymous calls for a few minutes
+```
+
+At `MLL-Lab/embodied-exam-runs@dfb89b3` the check passes for all 6,248 counted attempts (5,839 with a video).
+
+## How this maps to the website
+
+- Leaderboard page: `numbers.json` `per_agent` (Core and the three mode views).
+- Task pages: each task's Results row = `leaderboard_core.csv` (S = success, F = failed or missing).
+- "Watch the run" links and the `/run/` page: `viewer_url`, served from `MLL-Lab/embodied-exam-runs`.
+- A run whose last attempt an API or infrastructure error ended is shown with its verdict as counted and labelled
+  ("Ended by an API error"); its `run.json` has `ended_by`.

leaderboard/check_coverage.py

@@ -0,0 +1,53 @@
+"""Check that every counted leaderboard attempt has its run on HF (works without a token).
+
+usage: python3 check_coverage.py [trials.csv] [--revision REV]
+
+For each row of trials.csv (one row per counted attempt), it checks in the dataset MLL-Lab/embodied-exam-runs:
+  - runs/<h2>/<slug>/run.json and trajectory.jsonl.gz exist, and replay.mp4 exists when has_video = 1;
+  - the slug has a row in index/<suite>.json whose run and outcome equal the leaderboard's.
+It lists the runs/<h2>/ folders through the HF API (at most 256 calls) and reads the index files. Without a token,
+HF's rate limit may make it wait a few minutes; with HF_TOKEN set it runs in about a minute.
+Exit status 0 only at 100% coverage.
+"""
+import argparse, collections, csv, hashlib, json, sys, time, urllib.request
+from huggingface_hub import HfApi
+REPO = 'MLL-Lab/embodied-exam-runs'
+ap = argparse.ArgumentParser(); ap.add_argument('trials', nargs='?', default='trials.csv'); ap.add_argument('--revision', default='main')
+a = ap.parse_args()
+api = HfApi()
+rev = api.dataset_info(REPO, revision=a.revision).sha
+rows = list(csv.DictReader(open(a.trials)))
+h2 = lambda s: hashlib.sha1(s.encode()).hexdigest()[:2]
+files = set()
+def tree(d):   # anonymous calls are rate limited (a few hundred per 5 min): wait and retry
+    for k in range(20):
+        try:
+            return [e.path for e in api.list_repo_tree(REPO, path_in_repo=f'runs/{d}', recursive=True, repo_type='dataset', revision=rev)]
+        except Exception as e:
+            if '429' not in str(e) and 'rate limit' not in str(e).lower() or k == 19: raise
+            print('rate limited, waiting 60 s', file=sys.stderr, flush=True); time.sleep(60)
+for d in sorted({h2(r['hf_slug']) for r in rows}):
+    files.update(tree(d))
+def get(path):   # plain urllib: no compressed transfer
+    return json.loads(urllib.request.urlopen(f'https://huggingface.co/datasets/{REPO}/resolve/{rev}/{path}', timeout=300).read())
+idx = get('index.json')
+irow = {}
+for s in idx['suites']:
+    for x in get(f'index/{s}.json')['runs']:
+        irow[x['slug']] = x
+c = collections.Counter(); bad = []
+for r in rows:
+    s = r['hf_slug']; base = f'runs/{h2(s)}/{s}/'
+    need = ['run.json', 'trajectory.jsonl.gz'] + (['replay.mp4'] if r['has_video'] == '1' else [])
+    miss = [f for f in need if base + f not in files]
+    x = irow.get(s)
+    ok_idx = bool(x) and x['run'] == r['run_id'] and ((x.get('success') == 1 or x.get('state') == 'success') == (r['verdict'] == 'success'))
+    if miss or not ok_idx:
+        bad.append((r['run_id'], r['task_dir'], s, miss, 'index ok' if ok_idx else ('index outcome differs' if x else 'not in index')))
+    else:
+        c['covered'] += 1
+    c['with video'] += r['has_video'] == '1'
+print(f'{REPO}@{rev[:10]}: {c["covered"]} of {len(rows)} counted attempts covered '
+      f'(run.json + trajectory + index row with the same outcome; replay.mp4 for the {c["with video"]} with a video)')
+for b in bad: print('MISSING', *b)
+sys.exit(0 if not bad else 1)

leaderboard/numbers.json

@@ -0,0 +1,228 @@
+{
+ "snapshot": {
+  "taken_at": "2026-11-05T23:17:48-06:00",
+  "newest_counted_trial_finished_at": "2026-11-03T00:20:36Z"
+ },
+ "rule": "Task set: the exam repository main at ebd10b62b0 (1,156 tasks / 2,288 task dirs). Core: that commit's docs/core_set.toml (456 tasks / 901 tests; default modes 403 standard + 53 realtime). Per (agent run, test): the latest finished valid attempt (success or failed; infrastructure failures re-run, ablation and quick-check runs left out). Success = reward >= 1 within the token budget. A test with no counted attempt is missing and counts as failed. Core view = each core task's default test; Privileged / Standard / Realtime = every core test of that mode; rate = solved / tests of the view.",
+ "all": {
+  "tasks": 1156,
+  "dirs": 2288,
+  "dirs_per_mode": {
+   "privileged": 1047,
+   "standard": 1155,
+   "realtime": 86
+  },
+  "suites": 35
+ },
+ "core": {
+  "tasks": 456,
+  "tests": 901,
+  "tests_per_mode": {
+   "privileged": 391,
+   "standard": 456,
+   "realtime": 54
+  },
+  "defaults": {
+   "standard": 403,
+   "realtime": 53
+  },
+  "commit": "ebd10b62b0"
+ },
+ "views": {
+  "core": "each core task's default test",
+  "privileged": "every core privileged test",
+  "standard": "every core standard test",
+  "realtime": "every core realtime test"
+ },
+ "per_agent": {
+  "codex-gpt6_astra-high": {
+   "core": {
+    "solved": 150,
+    "tests": 456,
+    "rate": 0.328947,
+    "missing_as_fail": 2
+   },
+   "privileged": {
+    "solved": 303,
+    "tests": 391,
+    "rate": 0.774936,
+    "missing_as_fail": 1
+   },
+   "standard": {
+    "solved": 156,
+    "tests": 456,
+    "rate": 0.342105,
+    "missing_as_fail": 2
+   },
+   "realtime": {
+    "solved": 17,
+    "tests": 54,
+    "rate": 0.314815,
+    "missing_as_fail": 0
+   },
+   "missing_as_fail_total": 3
+  },
+  "claude_code-opus_5_5-high": {
+   "core": {
+    "solved": 107,
+    "tests": 456,
+    "rate": 0.234649,
+    "missing_as_fail": 0
+   },
+   "privileged": {
+    "solved": 231,
+    "tests": 391,
+    "rate": 0.590793,
+    "missing_as_fail": 1
+   },
+   "standard": {
+    "solved": 114,
+    "tests": 456,
+    "rate": 0.25,
+    "missing_as_fail": 0
+   },
+   "realtime": {
+    "solved": 4,
+    "tests": 54,
+    "rate": 0.074074,
+    "missing_as_fail": 0
+   },
+   "missing_as_fail_total": 1
+  },
+  "antigravity_cli-gemini_4_argon-high": {
+   "core": {
+    "solved": 89,
+    "tests": 456,
+    "rate": 0.195175,
+    "missing_as_fail": 0
+   },
+   "privileged": {
+    "solved": 164,
+    "tests": 391,
+    "rate": 0.419437,
+    "missing_as_fail": 0
+   },
+   "standard": {
+    "solved": 96,
+    "tests": 456,
+    "rate": 0.210526,
+    "missing_as_fail": 0
+   },
+   "realtime": {
+    "solved": 4,
+    "tests": 54,
+    "rate": 0.074074,
+    "missing_as_fail": 0
+   },
+   "missing_as_fail_total": 0
+  },
+  "antigravity_cli-gemini_3_8_flash-high": {
+   "core": {
+    "solved": 27,
+    "tests": 456,
+    "rate": 0.059211,
+    "missing_as_fail": 0
+   },
+   "privileged": {
+    "solved": 96,
+    "tests": 391,
+    "rate": 0.245524,
+    "missing_as_fail": 0
+   },
+   "standard": {
+    "solved": 27,
+    "tests": 456,
+    "rate": 0.059211,
+    "missing_as_fail": 0
+   },
+   "realtime": {
+    "solved": 1,
+    "tests": 54,
+    "rate": 0.018519,
+    "missing_as_fail": 0
+   },
+   "missing_as_fail_total": 0
+  },
+  "claude_code-qwen3_8_max-high": {
+   "core": {
+    "solved": 18,
+    "tests": 456,
+    "rate": 0.039474,
+    "missing_as_fail": 6
+   },
+   "privileged": {
+    "solved": 37,
+    "tests": 391,
+    "rate": 0.094629,
+    "missing_as_fail": 29
+   },
+   "standard": {
+    "solved": 18,
+    "tests": 456,
+    "rate": 0.039474,
+    "missing_as_fail": 17
+   },
+   "realtime": {
+    "solved": 1,
+    "tests": 54,
+    "rate": 0.018519,
+    "missing_as_fail": 2
+   },
+   "missing_as_fail_total": 48
+  },
+  "dsh-deepseek_v4_1_flash-high": {
+   "core": {
+    "solved": 12,
+    "tests": 456,
+    "rate": 0.026316,
+    "missing_as_fail": 0
+   },
+   "privileged": {
+    "solved": 69,
+    "tests": 391,
+    "rate": 0.176471,
+    "missing_as_fail": 0
+   },
+   "standard": {
+    "solved": 12,
+    "tests": 456,
+    "rate": 0.026316,
+    "missing_as_fail": 0
+   },
+   "realtime": {
+    "solved": 1,
+    "tests": 54,
+    "rate": 0.018519,
+    "missing_as_fail": 0
+   },
+   "missing_as_fail_total": 0
+  },
+  "claude_code-glm_5_3_flash-high": {
+   "core": {
+    "solved": 11,
+    "tests": 456,
+    "rate": 0.024123,
+    "missing_as_fail": 4
+   },
+   "privileged": {
+    "solved": 36,
+    "tests": 391,
+    "rate": 0.092072,
+    "missing_as_fail": 3
+   },
+   "standard": {
+    "solved": 11,
+    "tests": 456,
+    "rate": 0.024123,
+    "missing_as_fail": 4
+   },
+   "realtime": {
+    "solved": 0,
+    "tests": 54,
+    "rate": 0.0,
+    "missing_as_fail": 0
+   },
+   "missing_as_fail_total": 7
+  }
+ }
+}