File size: 2,967 Bytes
6087947 f4cb77b 6087947 | 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 | ---
license: apache-2.0
task_categories:
- question-answering
- text-generation
language:
- en
tags:
- agent
size_categories:
- n<1K
---
# GISA: A Benchmark for General Information-Seeking Assistant</h1>
## Benchmark Highlights
GISA is a benchmark for General Information-Seeking Assistants with 373 human-crafted queries that reflect real-world information needs. It includes both stable and live subsets, four structured answer formats (item, set, list, table), and complete human search trajectories for every query.
- **Diverse answer formats with deterministic evaluation.**
GISA uses four structured answer types (item, set, list, table) with strict matching metrics for reproducible evaluation, avoiding subjective LLM judging while preserving task diversity.
- **Unified deep + wide search capabilities.**
Tasks require both vertical reasoning and horizontal information aggregation across sources, evaluating long-horizon exploration and summarization in one benchmark.
- **Dynamic, anti-static evaluation.**
Queries are split into stable and live subsets; the live subset is periodically updated to reduce memorization and keep the benchmark challenging over time.
- **Process-level supervision via human trajectories.**
Full human search trajectories are provided for every query, serving as gold references for process reward modeling and imitation learning while validating task solvability.
## Evaluation
Please refer to our [GitHub](https://anonymous.4open.science/r/GISA/).
## Data Schema
#### 1. encrypted_question.jsonl
Each row contains:
- id (int): the ID of the question (it is **not** continuous)
- question (str): the question after encryption
- answer_type (str): the type of the answer, can be item, set, list, or table
- question_type (str): the type of the question, can be stable or live
- topic (str): the topic of the question, can be TV Shows \& Movies, Science \& Technology, Art, History, Sports, Music, Video Games, Geography, Politics, or Other
- canary (str): the password used for decryption
#### 2. answer/[id].csv
The file contains the answer corresponds to the question [id].
#### 3. trace/[id].json
The file conatins the human trajectory of the question [id], with the following keys:
- search (list): the queries issued by the annotator
- result (dict): the search result of each query
- click (list): the click behaviors made by the annotator
-
## Loading Method
```python
def derive_key(password: str, length: int) -> bytes:
hasher = hashlib.sha256()
hasher.update(password.encode())
key = hasher.digest()
return key * (length // len(key)) + key[: length % len(key)]
def decrypt(ciphertext_b64: str, password: str) -> str:
encrypted = base64.b64decode(ciphertext_b64)
key = derive_key(password, len(encrypted))
decrypted = bytes(a ^ b for a, b in zip(encrypted, key))
return decrypted.decode()
obj["question"] = decrypt(str(obj["question"]), str(obj["canary"]))
```
|