skills/bioprobench/kernel.py

• _require · function · L52-L66 — def _require(task_name: str)
• _ensure_nltk_resources · function · L69-L107 — def _ensure_nltk_resources(nltk)
• obtain · function · L76-L85 — def obtain(resource_path, package)
• extract_binary_answer · function · L113-L133 — def extract_binary_answer(generated_str)
• normalize_bool_label · function · L140-L161 — def normalize_bool_label(value)
• _mean_or_none · function · L164-L168 — def _mean_or_none(values)
• evaluate_step_reasoning_model · function · L177-L205 — def evaluate_step_reasoning_model(result_path)
• evaluate_correction_task · function · L209-L245 — def evaluate_correction_task(output_file_path)
• evaluate_protocolgen_model · function · L249-L388 — def evaluate_protocolgen_model(result_path): # Lazy dependency resolution: GEN is the only task needing the torch-scale # stack, and the only one needing NLTK corpora off the network.
• extract_text_response · function · L268-L278 — def extract_text_response(text)
• _gt_index_order · function · L392-L413 — def _gt_index_order(wrong_steps, correct_steps)
• evaluate_sorting_predictions · function · L416-L480 — def evaluate_sorting_predictions(output_file_path)
• _split_answer_and_confidence · function · L487-L520 — def _split_answer_and_confidence(content)
• evaluate_predictions · function · L523-L561 — def evaluate_predictions(output_file_path)
• run_bioprobench_eval · function · L567-L636 — def run_bioprobench_eval(task_name: str, response_file_path: str)
• _result_status · function · L639-L659 — def _result_status(metrics: dict) -> str
• _temporary_modules · function · L669-L682 — def _temporary_modules(mapping: dict)
• _load_script · function · L685-L720 — def _load_script(module_name: str, file_name: str, siblings=())
• _safe_path_component · function · L726-L729 — def _safe_path_component(value: str, fallback: str) -> str
• get_task_prompt · function · L732-L738 — def get_task_prompt(sample: dict, task_name: str) -> str
• run_llm_judge_evaluation · function · L741-L762 — def run_llm_judge_evaluation( sample: dict, api_key: str, base_url: str = "https://api.deepseek.com", model_name: str = "deepseek-chat", ) -> str
• trigger_batch_inference · function · L765-L836 — def trigger_batch_inference( task_name: str, test_file_path: str, mode: str = "API", model_name: str = "o3-mini", api_key: str = None, base_url: str = None, ) -> dict