GenBench CoCG QA Test Set

Held-out test split of GenBench's multi-hop genetic reasoning QA dataset (source: genbench-iitp/genbench-coding-qa), built for CoCG (Co-Evolving Confidence Graph) agent training/evaluation.

1632 items total across two splits, one per pipeline:

test_coding (1472 items, pipeline coding_variant)

task_type	count
coding_variant	32
conservation_reasoning	160
counterfactual	160
disease_reasoning	160
evidence_attribution	160
hallucination_detection	160
interaction_propagation	160
mechanistic_explanation	160
path_traversal	160
structural_effect	160

test_noncoding (160 items, pipeline noncoding_regulatory)

task_type	count
tissue_specific	160

Schema

Each item has:

• id, task_type, pipeline (coding_variant/noncoding_regulatory), difficulty
• question, answer, choices (MCQ options, when applicable)
• context -- either a templated chain narration, or (if llm_rewrite was applied) an LLM-rewritten fluent Step/Evidence/Interpretation/Conclusion narrative
• reasoning_chain -- the grounded, machine-checkable multi-hop path (steps: each with source_node_id/target_node_id/edge_relation/ edge_confidence/edge_source_db), never touched by any LLM step
• modality_data -- raw modality payloads (sequence, structural, transcriptomic, post_translational, signaling_role, etc.) attached to the chain's anchor nodes
• evidence -- supporting evidence entries with source database/PMID
• path_confidence_score -- continuous, confidence-derived difficulty score

Companion graph

graph.json is the exact knowledge graph these items' reasoning_chain node IDs refer to -- load it with GenBench's GraphBuilder.load() to resolve full node/edge attributes beyond what's inlined in each item.

Source

Test split of genbench-iitp/genbench-coding-qa, computed by scripts/split_qa_dataset.py in GenBench, stratified by (pipeline, task_type) so every task type is represented.