README.md

---
license: unknown
task_categories:
- question-answering
tags:
- genomics
- knowledge-graph
- multi-hop-reasoning
- biology
pretty_name: GenBench CoCG QA Dataset
dataset_info:
  features:
  - name: id
    dtype: string
  - name: task_type
    dtype: string
  - name: pipeline
    dtype: string
  - name: difficulty
    dtype: string
  - name: num_hops
    dtype: int64
  - name: question
    dtype: string
  - name: context
    dtype: string
  - name: reasoning_chain
    struct:
    - name: steps
      list:
      - name: hop
        dtype: int64
      - name: source_node_id
        dtype: string
      - name: source_node_type
        dtype: string
      - name: source_node_label
        dtype: string
      - name: edge_relation
        dtype: string
      - name: edge_evidence_type
        dtype: string
      - name: edge_confidence
        dtype: float64
      - name: edge_source_db
        dtype: string
      - name: target_node_id
        dtype: string
      - name: target_node_type
        dtype: string
      - name: target_node_label
        dtype: string
      - name: human_readable
        dtype: string
    - name: total_hops
      dtype: int64
    - name: start_node_id
      dtype: string
    - name: end_node_id
      dtype: string
  - name: answer
    dtype: string
  - name: modalities
    list: string
  - name: modality_data
    list:
    - name: family
      dtype: string
    - name: tag
      dtype: string
    - name: anchor_node_id
      dtype: string
    - name: source_db
      dtype: string
    - name: payload
      struct:
      - name: gene
        dtype: string
      - name: tissues
        list: string
      - name: tpm_median
        list: float64
      - name: n_tissues
        dtype: int64
      - name: top_tissue
        dtype: string
      - name: chrom
        dtype: string
      - name: center
        dtype: int64
      - name: start
        dtype: int64
      - name: end
        dtype: int64
      - name: width
        dtype: int64
      - name: sequence
        dtype: string
      - name: biosample
        dtype: string
      - name: marks
        struct:
        - name: H3K4me1
          dtype: float64
        - name: H3K4me3
          dtype: float64
        - name: H3K27ac
          dtype: float64
      - name: state
        dtype: string
    - name: hops
      list: int64
  - name: evidence
    list:
    - name: source_db
      dtype: string
    - name: source_id
      dtype: string
    - name: text
      dtype: string
    - name: confidence
      dtype: float64
  - name: metadata
    struct:
    - name: gene_a
      dtype: string
    - name: gene_b
      dtype: string
    - name: path_length
      dtype: int64
    - name: gene
      dtype: string
    - name: tissue
      dtype: string
    - name: element_type
      dtype: string
    - name: variant
      dtype: string
    - name: pathway
      dtype: string
    - name: mode
      dtype: string
    - name: fabricated
      dtype: string
    - name: real_options
      list: string
    - name: protein
      dtype: string
    - name: threshold
      dtype: int64
    - name: partner_count
      dtype: int64
    - name: disease
      dtype: string
    - name: pathogenic_variants
      list: string
    - name: top_tissue
      dtype: string
    - name: top_tpm
      dtype: float64
  - name: path_confidence_score
    dtype: float64
  splits:
  - name: train
    num_bytes: 19898486
    num_examples: 4900
  download_size: 17322981
  dataset_size: 19898486
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---

# GenBench CoCG QA Dataset

Multi-hop genetic reasoning QA items generated from GenBench's knowledge
graph (Ensembl, ClinVar, VEP, BioGRID, STRING, Reactome, UniProt, GO,
SIGNOR, OmniPath, KEGG, DisGeNET, OpenTargets, PubTator3, GTEx, and more),
built for CoCG (Co-Evolving Confidence Graph) agent training.

**1464 items across 6 task types.**

## Task types

| task_type | count |
|---|---|
| `disease_reasoning` | 244 |
| `hallucination_detection` | 244 |
| `interaction_propagation` | 244 |
| `mechanistic_explanation` | 244 |
| `path_traversal` | 244 |
| `regulatory_reasoning` | 244 |

## 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` (if included in this repo) is the exact knowledge graph these
items' `reasoning_chain` node IDs refer to -- load it with
[GenBench's `GraphBuilder.load()`](https://github.com/murk07/GenBench) to
resolve full node/edge attributes beyond what's inlined in each item.

## Source

Generated from `data\processed\noncoding_qa_dataset.jsonl` in [GenBench](https://github.com/murk07/GenBench),
the substrate for CoCG (Co-Evolving Confidence Graph) agent training --
per-edge, per-modality KG confidence that co-adapts with an RL policy
during training rather than treating the KG as a frozen oracle.