Spaces:
Running
Running
File size: 5,151 Bytes
671787b 88875f7 671787b 48661cd 671787b 48661cd 671787b 1b91307 671787b 1b91307 671787b 1b91307 671787b 1b91307 671787b 1b91307 48661cd 671787b 48661cd 671787b 48661cd | 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 | """ Agents for Task3 : Rule Checking for a function """
import json
import random as _random
from typing import Any, Dict, List
from server import Task3Environment
from env.schemas import Action, ActionType
from data.data_loader import load_contracts, get_function_by_name
# ─────────────────────────────────────────────────────────────────────────────
# Helpers
# ─────────────────────────────────────────────────────────────────────────────
def _parse_fn_list(result_text: str) -> List[str]:
"""Parse 'Functions in X: f1, f2, f3' into [f1, f2, f3]."""
if ": " in result_text:
return [f.strip() for f in result_text.split(": ", 1)[-1].split(", ") if f.strip()]
return []
# ─────────────────────────────────────────────────────────────────────────────
# Task 3 agents
# ─────────────────────────────────────────────────────────────────────────────
def oracle_t3(env: Task3Environment, seed: int, verbose: bool = False) -> Dict[str, Any]:
"""Submits exact target function → score = 1.0."""
r = env.reset(seed=seed)
obs = r.observation
fn_name = env.state().target_function
contract = obs.contract_name
if verbose:
prop = obs.extra.get("property_english", "")[:60]
print(f" {contract}.{fn_name}() \"{prop}\"")
env.step(Action(action_type=ActionType.GET_PROPERTY_SPECIFICATION))
env.step(Action(action_type=ActionType.LIST_FUNCTIONS))
result = env.step(Action(action_type=ActionType.SUBMIT_FUNCTION,
params={"function_name": fn_name}))
return {"seed": seed, "contract": contract, "target_function": fn_name,
"grader_score": result.reward.value}
def subfunction_t3(env: Task3Environment, seed: int, verbose: bool = False) -> Dict[str, Any]:
"""Submits the first partial-credit subfunction if one exists, else 'constructor'."""
r = env.reset(seed=seed)
obs = r.observation
contracts = load_contracts()
target_contract = {}
for c in contracts:
if c["contract_name"] == obs.contract_name:
target_contract = c
break
submit_name = "constructor"
if (target_contract and "call_graph" in target_contract and env.state().target_function in
target_contract["call_graph"] and target_contract["call_graph"][env.state().target_function]):
submit_name = target_contract["call_graph"][env.state().target_function][0]
result = env.step(Action(action_type=ActionType.SUBMIT_FUNCTION,
params={"function_name": submit_name}))
if verbose:
prop = obs.extra.get("property_english", "")[:60]
print(f" {obs.contract_name}.{env.state().target_function}() \"{prop}\"")
print(f" Submitting subfunction: {submit_name}")
print(f" Reward received: {result.reward.value}")
return {"seed": seed, "grader_score": result.reward.value, "submitted": submit_name}
def random_t3(env: Task3Environment, seed: int) -> Dict[str, Any]:
"""Genuine random agent: lists functions, picks one at random, submits.
With N functions per contract and 1 target, expected score ≈ 1/N ≈ 0.20–0.25.
Uses a seeded RNG for reproducibility.
"""
rng = _random.Random(seed ^ 0xCAFE1)
env.reset(seed=seed)
# Step 1: get function list (necessary to pick a real candidate)
s = env.step(Action(action_type=ActionType.LIST_FUNCTIONS))
fns = _parse_fn_list(s.observation.last_action_result or "")
if not fns:
fns = ["constructor"]
# Step 2: optionally do 1 cheap browse action (formalized or call_graph)
browse_options = [
(ActionType.GET_PROPERTY_SPECIFICATION, {}),
(ActionType.GET_CALL_GRAPH, {}),
]
at, params = rng.choice(browse_options)
env.step(Action(action_type=at, params=params))
# Step 3: submit a uniformly random function from the real list
chosen = rng.choice(fns)
result = env.step(Action(action_type=ActionType.SUBMIT_FUNCTION,
params={"function_name": chosen}))
return {"seed": seed, "grader_score": result.reward.value, "submitted": chosen}
def floor_t3(env: Task3Environment, seed: int) -> Dict[str, Any]:
"""Always submits 'constructor' → guaranteed score = 0.0."""
env.reset(seed=seed)
result = env.step(Action(action_type=ActionType.SUBMIT_FUNCTION,
params={"function_name": "constructor"}))
return {"seed": seed, "grader_score": 0.001} |