pyrrho-v1-nano-g3.2

Historical Pyrrho v1 release. Current v2 model: yafitzdev/pyrrho-v2-nano-g1.

pyrrho-v1-nano-g3.2

pyrrho-v1-nano-g3.2 is a small multitask RAG governance co-processor for anti-hallucination and retrieval-quality pipelines. It reads a user question plus retrieved source passages, then returns a calibrated evidence-state decision and auxiliary signals that fitz-sage can use before answer generation.

It is not an answer generator and not an open-world fact checker. It sits between retrieval and generation, or beside a retrieval package as a fast evidence quality layer. Compared with pyrrho-v1-nano-g3.1, this package adds V8.2 retrieval-control heads for retrieval action, gap type, answerability shape, retrieval modality, and an additional evidence_failure_severity scalar.

Governance Labels

Multitask Heads

Outputs

This is a custom multitask package, not a standard single-head AutoModelForSequenceClassification artifact. The recommended runtime is pyrrho.multitask_inference.PyrrhoMultiTaskPredictor from the pyrrho repository.

The predictor returns a structured object:

Example normalized output shape:

{
  "schema_version": "pyrrho_multitask_prediction_v1",
  "governance": {
    "raw_label": "TRUSTWORTHY",
    "final_label": "TRUSTWORTHY",
    "used_threshold_fallback": false,
    "threshold": 0.34,
    "confidence": 0.84,
    "probabilities": {
      "ABSTAIN": 0.08,
      "DISPUTED": 0.08,
      "TRUSTWORTHY": 0.84
    }
  },
  "query_contract": {
    "final_label": "structured_lookup"
  },
  "route": {
    "final_label": "economics_finance"
  },
  "taxonomy": {
    "final_label": "direct_answer"
  },
  "retrieval_action": {
    "final_label": "answer_now"
  },
  "scalars": {
    "evidence_sufficiency": 0.91,
    "query_evidence_alignment": 0.88,
    "answer_coverage": 0.86,
    "conflict_density": 0.08,
    "retrieval_retry_value": 0.12,
    "false_trustworthy_risk": 0.09,
    "evidence_failure_severity": 0.07
  }
}

The model does not generate answers, citations, source spans, retrieval results, or natural-language explanations. It classifies and scores the (query, retrieved_contexts) evidence state.

Intended Use

Use this model when a RAG or retrieval package needs fast local signals about:

• whether retrieved evidence is enough to answer,
• whether retrieved evidence conflicts,
• what kind of evidence the query needs before retrieval,
• which semantic/domain route the query belongs to,
• which fitz-gov support/failure pattern is active,
• what retrieval action and gap type the evidence state suggests,
• whether retrieval should retry, broaden, or escalate.

This model is not intended to write answers, verify facts outside the provided sources, replace a retriever, or replace human review in high-stakes settings.

Quick Start

Install the pyrrho package from the repository that contains this runtime, then load the package with the multitask predictor:

from huggingface_hub import snapshot_download

from pyrrho.multitask_inference import PyrrhoMultiTaskPredictor

MODEL_ID = "yafitzdev/pyrrho-v1-nano-g3.2"
PACKAGE_DIR = snapshot_download(MODEL_ID)

query = "Which quarterly report is relevant?"
contexts = [
    "The Q2 report lists revenue, churn, and roadmap changes.",
]

predictor = PyrrhoMultiTaskPredictor.from_pretrained(PACKAGE_DIR, device="cpu")
result = predictor.predict(query, contexts)

print(result["governance"]["final_label"])
print(result["query_contract"]["final_label"])
print(result["route"]["final_label"])
print(result["taxonomy"]["final_label"])
print(result["retrieval_action"]["final_label"])
print(result["gap_type"]["final_label"])
print(result["scalars"])

For local package testing:

python scripts/package_multitask_encoder.py verify --package-dir models/pyrrho-v1-nano-g3.2 --device cpu

Release Selection

• Seed: 1337
• TRUSTWORTHY threshold: 0.34
• Selection reason: Seed 1337 was selected because it has the best held-out governance result among the completed g3.2 seeds: 97.56% accuracy / 0.89% false-trustworthy at tau 0.34, while also having the strongest retrieval-modality held-out macro F1.

Held-Out Test Metrics

Three-seed headline from the local release summary:

Training Data

Trained on fitz-gov V8.2.0 rows with mandatory routing.query_contract and routing.retrieval_control fields. The release package records the local training config in training_config.yaml and detailed metrics in reports/summary.json.

Limitations

• This is a governance and routing co-processor, not a generator.
• The auxiliary heads are useful signals, not ground-truth explanations.
• Query-contract and route predictions are query-only and can be wrong when the user query is underspecified.
• Taxonomy and scalar outputs are trained on fitz-gov labels/signals and should be treated as decision-support metadata, not universal factual judgments.
• Retrieval modality is the weakest new head; sparse subclasses such as pdf_layout should be treated as hints, not authoritative decisions.
• The V8.2 retrieval-control labels are mechanically validated, but did not receive a separate independent blind-label QA pass before this local candidate package.
• The license is CC BY-NC 4.0. Commercial use requires a separate license.