---
type: "post"
id: "code-and-pipeline-available-at-github-com-ponhvoan-62f346"
slug: "code-and-pipeline-available-at-github-com-ponhvoan-62f346"
title: "Code and pipeline available at github.com/ponhvoan/TruthAnchor under Apache-2.0."
description: "TruthAnchor makes uncertainty calibration an inspectable pipeline rather than an unspecified post-processing step."
retrieval_nugget: "TruthAnchor makes uncertainty calibration an inspectable pipeline rather than an unspecified post-processing step."
published_at: "2026-10-20"
updated_at: "2026-10-20"
record_date: "2026-10-20"
date_kind: "published_at"
topics: ["github"]
source_urls: ["https://github.com/ponhvoan/TruthAnchor"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:code-and-pipeline-available-at-github-com-ponhvoan-62f346"
status: "published"
visuals: []
editorial_provenance: {"schema_version":"newruntime-editorial-copy-v1","content_status":"source_grounded_final","final_copy_sha256":"sha256:beaba7555d918eef38187e3b9fd66dc007ddc3dbec98d4b4dc15fa7600d7650e","reviewed_at":"2026-10-20T02:01:43.820Z","source_evidence_count":1,"verified_claim_count":3,"site_analysis_schema_version":"newruntime-site-analysis-v1","site_object_kind":"field_note","observed_fact_count":3,"implication_count":1,"watch_condition_count":1,"related_record_count":0}
analysis: {"schema_version":"newruntime-site-analysis-v1","object_kind":"field_note","thesis":"A calibration release is useful as a research artifact when its score-to-correctness mapping and evaluation path can be rerun, but code availability alone does not establish better calibration.","observed_facts":[{"text":"The repository is public and carries an Apache-2.0 license.","source_urls":["https://github.com/ponhvoan/TruthAnchor"]},{"text":"It lists TriviaQA, SciQ, and PopQA as supported benchmark datasets.","source_urls":["https://github.com/ponhvoan/TruthAnchor"]},{"text":"The README describes generation, uncertainty scoring, mapper training, held-out evaluation, and outputs including evaluation results and score files.","source_urls":["https://github.com/ponhvoan/TruthAnchor"]}],"mechanism":"A lightweight MLP learns a mapping from raw uncertainty scores to an estimated probability that a response is correct.","why_now":"Researchers evaluating LLM confidence can inspect and run a concrete calibration workflow rather than infer one from a high-level claim.","implications":["The repository lowers the practical barrier to testing score calibration on the listed benchmarks or custom labeled data, while leaving the value of the learned mapping to be measured."],"evidence_boundary":"The supplied GitHub page documents implementation scope and usage, but provides no numerical results establishing accuracy, generalization, or superiority to alternatives.","watch_conditions":["Look for reported held-out metrics across datasets and models, including whether calibration gains persist outside the benchmark setup."],"related_records":[],"new_branch_reason":"No prior New Runtime record about TruthAnchor was supplied; the available related context concerns a different pipeline-optimization approach, not this score-calibration artifact."}
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/posts/code-and-pipeline-available-at-github-com-ponhvoan-62f346/","markdown":"https://newruntime.com/posts/code-and-pipeline-available-at-github-com-ponhvoan-62f346.md","json":"https://newruntime.com/posts/code-and-pipeline-available-at-github-com-ponhvoan-62f346.json"}
---

# Code and pipeline available at github.com/ponhvoan/TruthAnchor under Apache-2.0.

## Retrieval answer

TruthAnchor makes uncertainty calibration an inspectable pipeline rather than an unspecified post-processing step.

TruthAnchor makes uncertainty calibration an inspectable pipeline rather than an unspecified post-processing step. The thesis is: A calibration release is useful as a research artifact when its score-to-correctness mapping and evaluation path can be rerun, but code availability alone does not establish better calibration. Its mechanism is: A lightweight MLP learns a mapping from raw uncertainty scores to an estimated probability that a response is correct. The repository describes an end-to-end path through response generation, uncertainty scoring, mapper training, and held-out evaluation.

The observed facts are: the repository is public and carries an Apache-2.0 license; it lists TriviaQA, SciQ, and PopQA as supported benchmark datasets. It also describes outputs including evaluation results and score files, plus a walkthrough notebook and custom score-label CSV input. Why now: Researchers evaluating LLM confidence can inspect and run a concrete calibration workflow rather than infer one from a high-level claim. The first implication is: The repository lowers the practical barrier to testing score calibration on the listed benchmarks or custom labeled data, while leaving the value of the learned mapping to be measured.

The evidence boundary is: The supplied GitHub page documents implementation scope and usage, but provides no numerical results establishing accuracy, generalization, or superiority to alternatives. The first watch condition is: Look for reported held-out metrics across datasets and models, including whether calibration gains persist outside the benchmark setup. This begins a distinct analytical record because no prior New Runtime record about TruthAnchor was supplied; the available related context concerns a different pipeline-optimization approach, not this score-calibration artifact.

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- topic hub: [Github](https://newruntime.com/topics/github.md): Explore the github topic hub.
- collection: [Field Notes](https://newruntime.com/posts.json): Open the latest editorial analysis.
- collection: [Source Ledger](https://newruntime.com/source-ledger.json): Inspect the public source evidence graph.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)