Spaces:
Running
Running
File size: 12,824 Bytes
3c19b11 c65a116 3c19b11 a593bad 1efaba3 3c19b11 bb01fd2 55a6f3e 3c19b11 cdb54bf 3c19b11 7990e30 3c19b11 cdb54bf 78b04f0 3c19b11 78b04f0 3c19b11 a593bad 1efaba3 3c19b11 c65a116 3c19b11 a593bad 1efaba3 3c19b11 bb01fd2 55a6f3e 3c19b11 cdb54bf 3c19b11 c65a116 3c19b11 a593bad 1efaba3 3c19b11 bb01fd2 55a6f3e 3c19b11 cdb54bf 3c19b11 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 | """Leaderboard entry schema for Web Agent Bench."""
from __future__ import annotations
import re
from dataclasses import dataclass, field
from typing import Any
def public_source_path(value: str | None) -> str | None:
"""Keep only portable relative paths; drop machine-local absolute paths."""
if not value:
return None
text = str(value).strip()
if not text:
return None
if text.startswith("\\\\") or (len(text) >= 2 and text[1] == ":"):
return None
if text.startswith("/") and not text.startswith("results/"):
return None
return text.replace("\\", "/")
def entry_id_for(model: str, harness: str) -> str:
safe = re.sub(r"[^\w.\-]+", "_", f"{model}__{harness}".strip())
return safe or "unknown"
def reconcile_success_metrics(entry: LeaderboardEntry) -> None:
"""Primary success rate is always passed_tasks / total_tasks (DAB)."""
section_passed = sum(section.passed for section in entry.sections.values())
section_total = sum(section.total for section in entry.sections.values())
if section_total > 0 and entry.total_tasks <= 0:
entry.total_tasks = section_total
entry.passed_tasks = section_passed
entry.failed_tasks = section_total - section_passed
if entry.total_tasks > 0:
entry.success_rate = entry.passed_tasks / entry.total_tasks
entry.failed_tasks = entry.total_tasks - entry.passed_tasks
@dataclass
class SectionStats:
section_id: str
label: str
success_rate: float | None
passed: int
total: int
tasks: dict[str, bool] = field(default_factory=dict)
@dataclass
class LeaderboardEntry:
model: str
harness: str
provider: str | None = None
mock: str | None = None
finished: bool = True
source_path: str | None = None
submitted_at: str | None = None
success_rate: float | None = None
passed_tasks: int = 0
failed_tasks: int = 0
total_tasks: int = 0
avg_duration_seconds: float | None = None
total_duration_seconds: float | None = None
total_agent_steps: int | None = None
avg_agent_steps: float | None = None
avg_tokens_per_task: float | None = None
total_tokens: int | None = None
total_cost_usd: float | None = None
avg_cost_per_task_usd: float | None = None
token_usage_by_model: dict[str, dict[str, Any]] = field(default_factory=dict)
ouroboros_model_slots: dict[str, list[str]] = field(default_factory=dict)
pass_at_k: float | None = None
agent_completion_rate: float | None = None
agent_dab_agreement_rate: float | None = None
strict_passed_tasks: int | None = None
sections: dict[str, SectionStats] = field(default_factory=dict)
ui_badges: list[str] = field(default_factory=list)
ui_classes: dict[str, dict[str, Any]] = field(default_factory=dict)
section_avg_rate: float | None = None
raw_results_path: str | None = None
@property
def entry_id(self) -> str:
return entry_id_for(self.model, self.harness)
@property
def success_pct(self) -> float | None:
if self.success_rate is None:
return None
return round(self.success_rate * 100, 2)
@property
def strict_success_rate(self) -> float | None:
if self.strict_passed_tasks is None or self.total_tasks <= 0:
return None
return self.strict_passed_tasks / self.total_tasks
@property
def strict_success_pct(self) -> float | None:
rate = self.strict_success_rate
if rate is None:
return None
return round(rate * 100, 2)
@property
def input_modality(self) -> str:
from src.input_modality import harness_input
return harness_input(self.model, self.harness)
@property
def full_bench_success_rate(self) -> float | None:
from src.bench_config import task_count
if self.total_tasks >= task_count():
return self.success_rate
return None
@property
def value_score(self) -> float | None:
if self.success_rate is None or not self.total_cost_usd or self.total_cost_usd <= 0:
return None
return self.success_rate / self.total_cost_usd
def metric(self, name: str) -> float | None:
if name.startswith("domain:"):
section_id = name.split(":", 1)[1]
section = self.sections.get(section_id)
return section.success_rate if section else None
if name.startswith("taxonomy:"):
class_id = name.split(":", 1)[1]
row = self.ui_classes.get(class_id)
if not row:
return None
rate = row.get("success_rate")
return float(rate) if isinstance(rate, (int, float)) else None
if name == "success_rate":
return self.success_rate
if name == "full_bench_success_rate":
return self.full_bench_success_rate
if name == "avg_duration_seconds":
return self.avg_duration_seconds
if name == "total_duration_seconds":
return self.total_duration_seconds
if name == "total_agent_steps":
return float(self.total_agent_steps) if self.total_agent_steps is not None else None
if name == "total_cost_usd":
return self.total_cost_usd
if name == "value_score":
return self.value_score
return getattr(self, name, None)
def to_json(self) -> dict[str, Any]:
return {
"model": self.model,
"harness": self.harness,
"provider": self.provider,
"mock": self.mock,
"finished": self.finished,
"source_path": public_source_path(self.source_path),
"submitted_at": self.submitted_at,
"metrics": {
"success_rate": self.success_rate,
"passed_tasks": self.passed_tasks,
"failed_tasks": self.failed_tasks,
"total_tasks": self.total_tasks,
"avg_duration_seconds": self.avg_duration_seconds,
"total_duration_seconds": self.total_duration_seconds,
"total_agent_steps": self.total_agent_steps,
"avg_agent_steps": self.avg_agent_steps,
"avg_tokens_per_task": self.avg_tokens_per_task,
"total_tokens": self.total_tokens,
"total_cost_usd": self.total_cost_usd,
"avg_cost_per_task_usd": self.avg_cost_per_task_usd,
"token_usage_by_model": self.token_usage_by_model,
"ouroboros_model_slots": self.ouroboros_model_slots,
"pass_at_k": self.pass_at_k,
"agent_completion_rate": self.agent_completion_rate,
"agent_dab_agreement_rate": self.agent_dab_agreement_rate,
"strict_passed_tasks": self.strict_passed_tasks,
"section_avg_rate": self.section_avg_rate,
},
"sections": {
section_id: {
"label": stats.label,
"success_rate": stats.success_rate,
"passed": stats.passed,
"total": stats.total,
"tasks": stats.tasks,
}
for section_id, stats in self.sections.items()
},
"ui_badges": self.ui_badges,
"ui_classes": self.ui_classes,
"section_avg_rate": self.section_avg_rate,
"raw_results_path": self.raw_results_path,
"entry_id": self.entry_id,
}
@classmethod
def from_json(cls, payload: dict[str, Any]) -> LeaderboardEntry:
if "metrics" in payload:
metrics = payload.get("metrics") or {}
sections_raw = payload.get("sections") or {}
sections = {
section_id: SectionStats(
section_id=section_id,
label=(data.get("label") or section_id),
success_rate=data.get("success_rate"),
passed=int(data.get("passed") or 0),
total=int(data.get("total") or 0),
tasks={k: bool(v) for k, v in (data.get("tasks") or {}).items()},
)
for section_id, data in sections_raw.items()
}
entry = cls(
model=str(payload.get("model") or "unknown"),
harness=str(payload.get("harness") or "unknown"),
provider=payload.get("provider"),
mock=payload.get("mock"),
finished=bool(payload.get("finished", True)),
source_path=public_source_path(payload.get("source_path")),
submitted_at=payload.get("submitted_at"),
success_rate=metrics.get("success_rate"),
passed_tasks=int(metrics.get("passed_tasks") or 0),
failed_tasks=int(metrics.get("failed_tasks") or 0),
total_tasks=int(metrics.get("total_tasks") or 0),
avg_duration_seconds=metrics.get("avg_duration_seconds"),
total_duration_seconds=metrics.get("total_duration_seconds"),
total_agent_steps=metrics.get("total_agent_steps"),
avg_agent_steps=metrics.get("avg_agent_steps"),
avg_tokens_per_task=metrics.get("avg_tokens_per_task"),
total_tokens=metrics.get("total_tokens"),
total_cost_usd=metrics.get("total_cost_usd"),
avg_cost_per_task_usd=metrics.get("avg_cost_per_task_usd"),
token_usage_by_model=dict(metrics.get("token_usage_by_model") or {}),
ouroboros_model_slots={
slot: [str(model) for model in models]
for slot, models in (metrics.get("ouroboros_model_slots") or {}).items()
if isinstance(models, list)
},
pass_at_k=metrics.get("pass_at_k"),
agent_completion_rate=metrics.get("agent_completion_rate"),
agent_dab_agreement_rate=metrics.get("agent_dab_agreement_rate"),
strict_passed_tasks=(
int(metrics["strict_passed_tasks"])
if isinstance(metrics.get("strict_passed_tasks"), int)
else None
),
sections=sections,
ui_badges=list(payload.get("ui_badges") or []),
ui_classes=dict(payload.get("ui_classes") or {}),
section_avg_rate=metrics.get("section_avg_rate") or payload.get("section_avg_rate"),
raw_results_path=payload.get("raw_results_path"),
)
reconcile_success_metrics(entry)
return entry
entry = from_legacy_payload(payload)
reconcile_success_metrics(entry)
return entry
def from_legacy_payload(payload: dict[str, Any]) -> LeaderboardEntry:
"""Convert pre-refactor LIBRA-style JSON."""
from src.bench_config import load_bench_config
config = load_bench_config()
sections: dict[str, SectionStats] = {}
for section_id, meta in config["sections"].items():
block = payload.get(section_id) or payload.get(meta["domain"]) or {}
if not isinstance(block, dict):
continue
tasks = {
task: bool(block.get(task, 0) >= 0.5)
for task in meta["tasks"]
if isinstance(block.get(task), (int, float))
}
passed = sum(1 for ok in tasks.values() if ok)
total = len(tasks)
domain_total = block.get("domain_total_score")
success_rate = float(domain_total) if isinstance(domain_total, (int, float)) else None
if success_rate is None and total:
success_rate = passed / total
sections[section_id] = SectionStats(
section_id=section_id,
label=meta["label"],
success_rate=success_rate,
passed=passed,
total=total,
tasks=tasks,
)
section_avg_rate = payload.get("total_score")
if not isinstance(section_avg_rate, (int, float)) and sections:
rates = [s.success_rate for s in sections.values() if s.success_rate is not None]
section_avg_rate = sum(rates) / len(rates) if rates else None
entry = LeaderboardEntry(
model=str(payload.get("model") or "unknown"),
harness=str(payload.get("harness") or "unknown"),
total_tasks=sum(section.total for section in sections.values()),
passed_tasks=sum(section.passed for section in sections.values()),
sections=sections,
section_avg_rate=float(section_avg_rate) if isinstance(section_avg_rate, (int, float)) else None,
)
reconcile_success_metrics(entry)
return entry
|