feat(dataset): add CHHallu-Src v1 — Chinese history hallucination & source-attribution benchmark#2525
Open
lizhuojunx86 wants to merge 1 commit into
Open
Conversation
…ource-attribution benchmark CHHallu-Src is a 400-question closed-book benchmark for historical-fact hallucination and source attribution over the pre-Qin to Three-Kingdoms Chinese classics. Answer keys are 100% derived from a public-domain primary-source corpus and adversarially validated. - Data (CC BY 4.0): https://huggingface.co/datasets/lizhuojun/chhallu-src-v1 - Tasks: sourcing (160) / cross-text (160) / attribution (80) - Completeness-robust cross-text metric: recall + era-impossibility - Adds opencompass/datasets/chhallu_src.py (BaseDataset + Evaluator, DATASET_SOURCE-aware), config chhallu_src_gen.py, __init__ export, and a dataset-index.yml entry. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Add CHHallu-Src v1 — a closed-book Chinese historical-fact hallucination & source-attribution benchmark
What it tests. 400 closed-book questions over 9 public-domain transmitted classics
(Shiji, Hanshu, Hou Hanshu, Sanguozhi, Zuozhuan, Lüshi Chunqiu, Zizhi Tongjian, Lunyu,
Mengzi), in three tasks:
Why it's sound. Answer keys are 100% derived from a public-domain primary-source
corpus (no LLM-authored keys) and adversarially validated by an independent LLM panel.
Data. https://huggingface.co/datasets/lizhuojun/chhallu-src-v1 (CC BY 4.0). The loader
downloads from the Hub automatically and honours
DATASET_SOURCE.Finding (6-model panel). Frontier models still cite chronologically-impossible sources
— 167 such over-predictions across the panel (DeepSeek 6/15 vs gemini-3.5-flash 64, the
latter despite perfect source recall).
Changes
opencompass/datasets/chhallu_src.py—CHHalluSrcDataset(BaseDataset)+CHHalluSrcEvaluator(BaseEvaluator)opencompass/configs/datasets/chhallu_src/chhallu_src_gen.py— generation configopencompass/datasets/__init__.py— exportdataset-index.yml— index entryAll additions; no existing files' logic modified.
Author: Li Zhuojun, Beijing Yubei Technology Co., Ltd. (info@zhuojun.li)