File size: 9,894 Bytes
b9c2fa3 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 | """Reference model implementations for baselines"""
import random
import subprocess
import tempfile
import time
from typing import List, Dict, Any, Optional
from dataclasses import dataclass
from abc import ABC, abstractmethod
from rockman.benchmark.schema import Problem, TaskType
from rockman.benchmark.runners import get_runner
class BaselineModel(ABC):
"""Abstract baseline model."""
@property
@abstractmethod
def name(self) -> str:
pass
@property
@abstractmethod
def version(self) -> str:
pass
@abstractmethod
def solve(self, problem: Problem) -> List[str]:
"""Generate completions for a problem. Returns list of samples."""
pass
class RandomBaseline(BaselineModel):
"""Random code generation baseline."""
@property
def name(self) -> str:
return "random"
@property
def version(self) -> str:
return "1.0"
def solve(self, problem: Problem) -> List[str]:
"""Generate random syntactically valid but semantically random code."""
# Extract function signature from prompt
import re
sig_match = re.search(r'(def\s+\w+\s*\([^)]*\)\s*(?:->\s*\w+\s*)?:)', problem.prompt)
if not sig_match:
return ["pass"]
signature = sig_match.group(1)
return_type = "int"
if "->" in signature:
return_type = signature.split("->")[1].split(":")[0].strip()
# Generate random implementations
templates = {
"int": [
"return 0",
"return 42",
"return -1",
"return len([])",
"return sum([])",
],
"bool": [
"return True",
"return False",
"return True if False else False",
],
"str": [
'return ""',
'return "hello"',
'return str(0)',
],
"list": [
"return []",
"return [1, 2, 3]",
"return list(range(10))",
],
"dict": [
"return {}",
"return {'a': 1}",
],
}
implementations = templates.get(return_type, ["pass"])
return [f"{signature}\n {impl}" for impl in implementations[:5]]
class HeuristicBaseline(BaselineModel):
"""Simple heuristic-based baseline (e.g., return first input, max, min, etc.)."""
@property
def name(self) -> str:
return "heuristic"
@property
def version(self) -> str:
return "1.0"
def solve(self, problem: Problem) -> List[str]:
import re
sig_match = re.search(r'(def\s+\w+\s*\([^)]*\)\s*(?:->\s*\w+\s*)?:)', problem.prompt)
if not sig_match:
return ["pass"]
signature = sig_match.group(1)
params = re.search(r'\(([^)]*)\)', signature)
param_names = [p.split(":")[0].strip() for p in params.group(1).split(",")] if params else []
# Heuristics based on problem category/tags
category = problem.category.lower()
tags = [t.lower() for t in problem.tags]
if "sort" in category or "sort" in tags:
impl = "return sorted(args[0]) if args else []"
elif "max" in category or "maximum" in category:
impl = "return max(args[0]) if args and args[0] else 0"
elif "min" in category or "minimum" in category:
impl = "return min(args[0]) if args and args[0] else 0"
elif "sum" in category or "add" in category:
impl = "return sum(args) if args else 0"
elif "search" in category or "find" in category:
impl = "return args[0].index(args[1]) if len(args) > 1 and args[1] in args[0] else -1"
elif len(param_names) >= 2:
impl = f"return {param_names[0]}"
elif len(param_names) == 1:
impl = f"return {param_names[0]}"
else:
impl = "return 0"
return [f"{signature}\n {impl}"]
class TemplateBaseline(BaselineModel):
"""Template-based baseline for common patterns."""
@property
def name(self) -> str:
return "template"
@property
def version(self) -> str:
return "1.0"
TEMPLATES = {
("binary search", "search"): """
left, right = 0, len(arr) - 1
while left <= right:
mid = (left + right) // 2
if arr[mid] == target:
return mid
elif arr[mid] < target:
left = mid + 1
else:
right = mid - 1
return -1
""",
("two pointers", "two-pointer"): """
left, right = 0, len(arr) - 1
while left < right:
current = arr[left] + arr[right]
if current == target:
return [left, right]
elif current < target:
left += 1
else:
right -= 1
return []
""",
("sliding window", "sliding-window"): """
max_sum = float('-inf')
current_sum = 0
for i, val in enumerate(arr):
current_sum += val
if i >= k:
current_sum -= arr[i - k]
if i >= k - 1:
max_sum = max(max_sum, current_sum)
return max_sum
""",
("dp", "dynamic-programming"): """
n = len(arr)
dp = [0] * (n + 1)
for i in range(n):
dp[i + 1] = max(dp[i], dp[i] + arr[i]) # Placeholder
return dp[n]
""",
("bfs", "graph"): """
from collections import deque
visited = set([start])
queue = deque([(start, 0)])
while queue:
node, dist = queue.popleft()
if node == target:
return dist
for neighbor in graph[node]:
if neighbor not in visited:
visited.add(neighbor)
queue.append((neighbor, dist + 1))
return -1
""",
}
def solve(self, problem: Problem) -> List[str]:
import re
sig_match = re.search(r'(def\s+\w+\s*\([^)]*\)\s*(?:->\s*\w+\s*)?:)', problem.prompt)
if not sig_match:
return ["pass"]
signature = sig_match.group(1)
prompt_lower = problem.prompt.lower()
category = problem.category.lower()
tags = [t.lower() for t in problem.tags]
# Match template
for keywords, template in self.TEMPLATES.items():
if any(kw in prompt_lower or kw in category or kw in tags for kw in keywords):
return [f"{signature}\n{template}"]
# Default based on category
defaults = {
"graph": "return []",
"tree": "return None",
"dynamic programming": "return 0",
"greedy": "return 0",
"math": "return 0",
}
impl = defaults.get(category, "pass")
return [f"{signature}\n {impl}"]
class LLMBaseline(BaselineModel):
"""Baseline using an LLM API (placeholder for actual integration)."""
def __init__(self, model_name: str, api_key: str = None, temperature: float = 0.2):
self._model_name = model_name
self.api_key = api_key
self.temperature = temperature
@property
def name(self) -> str:
return f"llm-{self._model_name}"
@property
def version(self) -> str:
return "1.0"
def solve(self, problem: Problem) -> List[str]:
# Placeholder - would integrate with actual LLM API
# For now, return heuristic baseline
heuristic = HeuristicBaseline()
return heuristic.solve(problem)
def run_baseline_suite(
problems: List[Problem],
baselines: List[BaselineModel] = None,
samples_per_problem: int = 5
) -> Dict[str, Dict[str, List[str]]]:
"""Run multiple baselines on a problem set."""
if baselines is None:
baselines = [RandomBaseline(), HeuristicBaseline(), TemplateBaseline()]
all_completions = {}
for baseline in baselines:
print(f"Running baseline: {baseline.name} v{baseline.version}")
completions = {}
for problem in problems:
try:
samples = baseline.solve(problem)
# Pad or truncate to samples_per_problem
while len(samples) < samples_per_problem:
samples.append(samples[0] if samples else "pass")
completions[problem.task_id] = samples[:samples_per_problem]
except Exception as e:
print(f" Error on {problem.task_id}: {e}")
completions[problem.task_id] = ["pass"] * samples_per_problem
all_completions[baseline.name] = completions
return all_completions
def evaluate_baselines(
problems: List[Problem],
baseline_completions: Dict[str, Dict[str, List[str]]],
k: int = 1,
include_hidden: bool = False
) -> Dict[str, Any]:
"""Evaluate all baseline completions."""
from rockman.benchmark.evaluator import Evaluator
from rockman.benchmark.metrics import aggregate_scores
evaluator = Evaluator()
results = {}
for baseline_name, completions in baseline_completions.items():
print(f"\nEvaluating {baseline_name}...")
eval_result = evaluator.evaluate_dataset(problems, completions, include_hidden, k)
results[baseline_name] = eval_result
return results
def print_baseline_comparison(results: Dict[str, Any], problems: List[Problem]):
"""Print comparison of baseline results."""
from rockman.benchmark.metrics import aggregate_scores
print(f"\n{'='*80}")
print("Baseline Comparison")
print(f"{'='*80}")
print(f"{'Baseline':<20} {'Pass@1':<10} {'Evaluated':<12}")
print("-" * 50)
for baseline_name, result in results.items():
pass_at_k = result.get("mean_pass_at_k", 0)
evaluated = result.get("evaluated_count", 0)
print(f"{baseline_name:<20} {pass_at_k:.4f} {evaluated:<12}")
print(f"{'='*80}") |