Polyweave
Repositories
1
 Projects Packages

@polyweave/uncertainty (1.0.2)

Published 2026-07-10 21:00:45 +00:00 by Dvorak

Installation

@polyweave:registry=https://hub.kl1.tenere.ai/api/packages/Polyweave/npm/
npm install @polyweave/uncertainty@1.0.2
"@polyweave/uncertainty": "1.0.2"

About this package

Uncertainty

Uncertainty metrics for LLM outputs: radial dispersion, semantic entropy, ensemble consistency, and hallucination detection.

Synopsis

import { radialDispersionScore, semanticEntropy, ensembleConsistency } from '@polyweave/uncertainty';

const samples = [
  { text: 'Paris is the capital of France.', embedding: [0.1, 0.2, 0.3] },
  { text: 'The capital of France is Paris.', embedding: [0.12, 0.19, 0.31] },
  { text: 'France has Paris as its capital.', embedding: [0.11, 0.21, 0.29] }
];

const rds = radialDispersionScore(samples);
console.log(rds); // low score = high agreement

const entropy = semanticEntropy(samples);
console.log(entropy); // low entropy = high confidence

Install

npm install @polyweave/pool

Description

Uncertainty provides metrics for quantifying LLM output confidence. It supports:

  • Radial dispersion: embedding-space dispersion of parallel samples
  • Semantic entropy: clustering-based response entropy
  • Ensemble consistency: inter-sample agreement scoring
  • Logprob metrics: token-level probability analysis
  • Token signals: per-token uncertainty features
  • Prompt perturbation: sensitivity to prompt variations
  • Multi-turn confidence: confidence across conversation turns

API

radialDispersionScore(samples)

Compute radial dispersion in embedding space. Lower score = higher agreement.

semanticEntropy(samples, options)

Cluster samples and compute entropy over response clusters.

ensembleConsistency(samples, options)

Score agreement between multiple samples.

Logprob Metrics

logprobScore, logprobRatio, averagePointwiseEntropy, tokenEntropy

Semantic Metrics

textSimilarity, charOverlap

Token Signals

computeTokenSignals, extractBinaryChoiceConfidence

SESE (Semantic Entropy)

computeSESEScore, computeDirectedSESEScore

HHEM (Hierarchical Hallucination Ensemble Metric)

computeHHEMScore

Prompt Perturbation

computePerturbationScore, computeResponseSimilarity

Multi-Turn Confidence

computeMultiTurnConfidence, buildConfidencePrompt

Citation Offsets

resolveCitationOffsets, computeCitationCertainty, computeCitationUncertainty

Tests

npm test

Tests cover all metrics with expected values, the uncertainty wrapper duplex, PARAM configuration, END-triggered sampling, error handling, and emit-all mode.

Requirements

Node.js 18 or later. ESM only.

License

MIT

Dependencies

Development Dependencies

ID Version
@polyweave/core *
@rigor/core *

Peer Dependencies

ID Version
@polyweave/core *

Keywords

polyweave uncertainty logprobs entropy semantic hallucination radial-dispersion confidence
Details
npm
2026-07-10 21:00:45 +00:00
363
John Dvorak
SEE LICENSE IN LICENSE
27 KiB
Assets (1)
uncertainty-1.0.2.tgz 27 KiB
Versions (4) View all
1.0.7 2026-07-11
1.0.4 2026-07-11
1.0.3 2026-07-11
1.0.2 2026-07-10