SPXI instrument · method 0.9 of 87, level 0 (Measurement): Capture Registry (longitudinal observation). An instrument of SPXI, the Semantic Packet for eXchange & Indexing; the whole practice is itemized at spxi.dev/methods.

Capture Registry › capture olmotrace-aio-20261029

One record of the canonical Capture Registry (EA-WG-CAPTURES-01), cited at https://www.alexanarch.org/captures/olmotrace-aio-20261029/. the canonical Capture Registry (version 12.379) · the address page · this card in the gallery · this record as data · table of contents.

AXN:0AD3.CAPTURE — this address. Each observation below is one AXN, sealed with its analysis and measures; the identifier names the reading together with the response it reads.
Machine Reception2026-10-29
OLMoTrace
Originator: Allen Institute for AI (Ai2), with the University of Washington, UC Berkeley and Stanford · outside the archive (out-universe) · service · SPXI treatment: partial
BROAD MATCH
no
image
OLMOTRACE, BY NAME, ON GOOGLE'S AI OVERVIEW, SIGNED OUT, 2026-10-29: 'an open-source system developed by the Allen Institute for Artificial Intelligence (Ai2) that traces language model outputs back to their multi-trillion-token training datasets in real time.' How it works (the Show OLMoTrace button highlights verbatim spans; matches against corpora such as Dolma, ranking up to 10 documents by BM25; an extended infini-gram at O(L log N)), supported models (OLMo 2 32B and 13B Instruct, OLMoE 1B 7B, the OLMo 3 family), uses (fact-checking, hallucination diagnosis, auditing). Six cards: Ai2's blog, arXiv 2504.07096, the ACL Anthology, Ai2 on X (OLMo 3), Techzine, Ken Yeung's newsletter.
Full record — 6,686 characters, 6 sources
Capture record
captured
2026-10-29
surface
Google AI Overview
auth state
signed out, incognito
evidence class
paste
PER (historical scalar)
0.25
PER units retained
inst, id, src
citations read
6
observation id
OBS-3fd20ae244c0
address id
ADDR-435813c322c8
Reading

The innocent reading first. The Overview names the tool and its maker, explains how it finds training text in a model's answer, lists the models it covers and what it is good for.

Findings
Analysis analyst prose, not machine text

A provenance tool whose authors state precisely what a match does not mean is composed with that meaning restored. The composition turns a trace of verbatim overlap into evidence for a claim, the conversion of a matching relation into a citation relation that the paper names and refuses.

Machine text, verbatim
CAPTURE-TIME VERBATIM RECORD — paste of the Overview expanded, with its source cards · complete as supplied · READ IN FULL 2026-10-29
[Google AI Overview at the query, signed out and incognito (operator defaults of 2026-10-01 and 2026-10-14), 2026-10-29. The paste is the Overview expanded (the 'AI Mode Conversation' view, with the source cards that follow the answer); the 'You said' line carries the query twice, run together, and is given once here. Inline links are Google redirect URLs, kept as pasted.] [QUERENT] OLMoTrace [SYSTEM] [OLMoTrace](https://www.google.com/goto?url=CAESWgHrOzAVoVYHLgJxMVIjxyaZx7wDArKB6jxCb1tOprJFofLF9_o_Lr6foJNzJ2skLWhN04W7o7uFmFuQkrV4xPmu8JuLWWjcOIm7uniqJ_3AgD9QQJlTrCUJTQ) is an open-source system developed by the Allen Institute for Artificial Intelligence (Ai2) that traces language model outputs back to their multi-trillion-token training datasets in real time. [[1], [2]] How It Works * Real-Time Tracing: After a model generates a response in the Ai2 Playground, users can click the Show OLMoTrace button to highlight verbatim spans of text. [[1], [2]] * Source Retrieval: The tool matches output segments against training corpora (such as the Dolma datasets) and ranks up to 10 context documents using BM25 and unigram probabilities. [[1], [2]] * Underlying Tech: It uses an extended version of infini-gram with a parallel algorithm that reduces lookup complexity to \(O(L \log N)\). [1] Supported Models * OLMo 2 32B Instruct * OLMo 2 13B Instruct * OLMoE 1B 7B Instruct * OLMo 3 family (including OLMo 3-Think 32B) [[1], [2]] Key Use Cases * Fact-Checking: Verify whether specific claims in a model's answer are backed by real training documents or generated from general parameters. [[1], [2]] * Hallucination Diagnosis: Spot where a model mixes up facts or memorized text from source files. [[1], [2]] * Auditing & Transparency: Inspect the data recipes and pretraining paths for regulatory compliance or research. [[1], [2]] Would you like to know how to access the open-source code for OLMoTrace on GitHub/Hugging Face, or are you looking for details on a specific OLMo model version? Allen Institute for Artificial Intelligence
Sources (6) as cited, and as the copy produced them
  1. Allen Institute for Artificial Intelligence third_party
    Going beyond open data – increasing transparency and trust in language models with OLMoTrace | Ai2
    Feature: OLMoTrace in Ai2 Playground traces language model outputs to multi-trillion-token training data in real time. Supported Models: OLMo 2 32B Instruct, OL...
    as pastedGoing beyond open data – increasing transparency and trust in language models with OLMoTrace | Ai2 Feature: OLMoTrace in Ai2 Playground traces language model outputs to multi-trillion-token training data in real time. Supported Models: OLMo 2 32B Instruct, OL... arXiv
  2. arXiv third_party
    OLMoTrace: Tracing Language Model Outputs Back to Trillions of Training ...
    [2504.07096] OLMoTrace: Tracing Language Model Outputs Back to Trillions of Training Tokens Scheduled maintenance: Wednesday, October 28, 7:00 to 12:00 EDT (11:
    as pastedOLMoTrace: Tracing Language Model Outputs Back to Trillions of Training ... [2504.07096] OLMoTrace:
  3. ACL Anthology third_party
    Tracing Language Model Outputs Back to Trillions of Training Tokens
    We present OLMOTRACE, the first system that traces the outputs of language models back to their full, multi-trillion-token training data in real time. OLMOTRACE...
    as pastedTracing Language Model Outputs Back to Trillions of Training Tokens Scheduled maintenance: Wednesday, October 28, 7:00 to 12:00 EDT (11: ACL Anthology Tracing Language Model Outputs Back to Trillions of Training Tokens We present OLMOTRACE, the first system that traces the outputs of language models back to their full, multi-trillion-token training data in real time. OLMOTRACE... X
  4. X third_party
    Olmo 3 is wired into OlmoTrace in the Ai2 Playground, so you don't just ... - X
    Ai2 on X: "Olmo 3 is wired into OlmoTrace in the Ai2 Playground, so you don't just see its behavior—you can trace it. For example, you can ask Olmo 3-Think (32B...
    as pastedOlmo 3 is wired into OlmoTrace in the Ai2 Playground, so you don't just ... - X Ai2 on X: "Olmo 3 is wired into OlmoTrace in the Ai2 Playground, so you don't just see its behavior—you can trace it. For example, you can ask Olmo 3-Think (32B... Techzine Global
  5. Techzine Global third_party
    Ai2 introduces OLMoTrace: a peek inside LLM's - Techzine Global
    After generating an answer from OLMo in the Ai2 Playground, users can activate the tool by clicking the “Show OLMoTrace” button below the output. The tool then ...
    as pastedAi2 introduces OLMoTrace: a peek inside LLM's - Techzine Global After generating an answer from OLMo in the Ai2 Playground, users can activate the tool by clicking the “Show OLMoTrace” button below the output. The tool then ... The AI Economy | Ken Yeung
  6. The AI Economy | Ken Yeung third_party
    Meet OLMoTrace: The Tool That Shows Where AI Gets Its Info
    Ken Yeung Apr 09, 2025 1 Ai2 has launched a tool to help developers answer a critical question: Where do AI models get their data? Called OLMoTrace, it's an ope...
    as pastedMeet OLMoTrace: The Tool That Shows Where AI Gets Its Info Ken Yeung Apr 09, 2025 1 Ai2 has launched a tool to help developers answer a critical question: Where do AI models get their data? Called OLMoTrace, it's an ope...
↻ Re-runpermalink