Testing OpenAI’s o1, a model trained to reason step by step, the authors of “Embers of Autoregression” found that it substantially outperformed earlier models “in many cases” but “still displays the same qualitative trends”: its accuracy at decoding a shift cipher ranged from 47% in the lowest-probability case to 92% in the highest, and it used more reasoning tokens on low-probability examples.

SocioLens project,

Claim LD01-SL-C37 (v1.0)

Research opportunities this claim motivates

What this claim leaves open for research. Each is a hypothesis about a gap, not a finding.

Do the failure patterns of 2023–2024 language models persist in reasoning models?OpenAI's o1 shows the same pattern at higher accuracy (47% to 92%).

Provenance and status

Claim type: Estimate
Publication status: Eligible for publication, with qualifiers: preprint
Verification state: Checked against the primary source; the source is a preprint. Who checked this claim is not recorded for it; claim checks so far were AI-led, and no person has checked claims against their sources.
Does not license: A claim about later reasoning models, or that o1 cannot do these tasks
Limitations: Read in arXiv version 2 (4 October 2024); o1’s workings are not public; ceiling effects limit some comparisons, as the authors note; one author works for the developer of the model tested; Anthropic’s Claude 3, made by the developer of the model that drafted this debate, is among the models compared
Source (as cited): McCoy, Yao, Friedman, Hardy & Griffiths, “When a language model is optimized for reasoning, does it still show embers of autoregression? An analysis of OpenAI o1,” arXiv:2410.01792, 2024
Source type: Preprint (recorded as “Preprint”)
Institutional position: Mixed: independent academic (Yale University, Princeton University), Roundtable, and AI developer (one author at OpenAI, whose model is tested; he joined after the original project was completed)
Maturity of the evidence: Emerging evidence
Recorded date: 16 October 2026
Currentness / correction: This is the current version. No correction or concern is recorded on its sources.

Sources this version cites

McCoy, Yao, Friedman, Hardy & Griffiths, “When a language model is optimized for reasoning, does it still show embers of autoregression? An analysis of OpenAI o1,” arXiv:2410.01792, 2024 (Preprint)
Read the original

Version

This is the current version, v1.0. Earlier versions stay available, and their records never change.

Versions: v1.0 · current

Raw record: every recorded field, with its coded values

All 25 canonical Claim fields, verbatim, with the coded values derived from them. The plain-language labels above are read from these.

Claim type (coded): ESTIMATE (as printed: “ESTIMATE”)
Publication status (coded): ELIGIBLE (PREPRINT) (as printed: “ELIGIBLE WITH PREPRINT QUALIFIER”)
Verification state (coded): CONTENT_VERIFIED_PRIMARY (PREPRINT)
Source type (coded): PREPRINT
Maturity (coded): EMERGING

#	Canonical field	Canonical value (verbatim)	Normalized (Claim Value Vocabulary v1.0.2)
1	CLAIM ID	LD01-SL-C37	—
2	PUBLIC-FACING CLAIM	Testing OpenAI’s o1, a model trained to reason step by step, the authors of “Embers of Autoregression” found that it substantially outperformed earlier models “in many cases” but “still displays the same qualitative trends”: its accuracy at decoding a shift cipher ranged from 47% in the lowest-probability case to 92% in the highest, and it used more reasoning tokens on low-probability examples.	—
3	CLAIM TYPE	ESTIMATE	base_types: ESTIMATE
4	RESEARCH STRAND	B — Evidence that could tell the accounts apart	—
5	MATURITY	EMERGING	states: EMERGING
6	SOURCE	McCoy, Yao, Friedman, Hardy & Griffiths, “When a language model is optimized for reasoning, does it still show embers of autoregression? An analysis of OpenAI o1,” arXiv:2410.01792, 2024	source_ids: SRC-SL-MCCOY-ET-AL-O1-ARXIV-2410-01792
7	SOURCE TYPE	Preprint	classes: PREPRINT
8	INSTITUTIONAL POSITION	Mixed: independent academic (Yale University, Princeton University), Roundtable, and AI developer (one author at OpenAI, whose model is tested; he joined after the original project was completed)	—
9	SCOPE	Deterministic text tasks in English; OpenAI o1 compared with earlier models, 2024	—
10	DESIGN	Re-test of earlier tasks varying output probability and task frequency, with accuracy and the number of reasoning tokens as measures	—
11	SAMPLE / DATASET	Tasks from the authors’ 2023 study, including shift ciphers, Pig Latin, article swapping and list reversal	—
12	MODEL / TOOL / VERSION	OpenAI o1 (preview, 2024) and earlier models	—
13	COMPARATOR	High- and low-probability versions of the same task; earlier models	—
14	MAIN RESULT	Shift-cipher accuracy 47% (lowest probability) to 92% (highest); large gains on rare task variants, but sensitivity to probability persists; more tokens used where probability is low	—
15	INFERENTIAL TARGET	Whether a model optimized for reasoning still shows traces of next-word prediction	—
16	DOES NOT LICENSE	A claim about later reasoning models, or that o1 cannot do these tasks	—
17	VERIFICATION STATE	CONTENT VERIFIED PREPRINT	state: CONTENT_VERIFIED_PRIMARY · qualifiers: PREPRINT · verified_by: UNKNOWN
18	LIMITATIONS	Read in arXiv version 2 (4 October 2024); o1’s workings are not public; ceiling effects limit some comparisons, as the authors note; one author works for the developer of the model tested; Anthropic’s Claude 3, made by the developer of the model that drafted this debate, is among the models compared	—
19	CONTRADICTORY / SUPPORTING SOURCES	The original study: LD01-SL-C18	—
20	GEOGRAPHY / LANGUAGE	English task set	—
21	DISTRIBUTIONAL LEDGER	Not analysed in this claim	—
22	R01 / FIELD-MAP LINKS	LD-01, T4, D3	—
23	WHAT WOULD CHANGE THIS CLAIM?	Reasoning models whose accuracy no longer depends on output probability	—
24	PUBLICATION STATUS	ELIGIBLE WITH PREPRINT QUALIFIER	state: ELIGIBLE · qualifiers: PREPRINT
25	RECORDED DATE	2026-10-16	—

Field 22 references resolved against Field Map v0.2: LD-01 (LIVING_DEBATE); T4 (TENSION); D3 (DOMAIN)

The sources as recorded, with their record IDs and coded source types:

Pinned Sources (field 6, version-specific)

“McCoy, Yao, Friedman, Hardy & Griffiths, “When a language model is optimized for reasoning, does it still show embers of autoregression? An analysis of OpenAI o1,” arXiv:2410.01792, 2024” → McCoy et al. (2024), “When a language model is optimized…” (SRC-SL-MCCOY-ET-AL-O1-ARXIV-2410-01792--v1.0) — McCoy, Yao, Friedman, Hardy & Griffiths, “When a language model is optimized for reasoning, does it still show embers of autoregression? An analysis of OpenAI o1,” arXiv:2410.01792, 2024; class PREPRINT (explicit public provenance pin)

Cited by Living Debate versions

What kind of thing is generative AI, and does the answer predict anything? (v0.1) LD-01

Cite this page

SocioLens project. (2026, October 16). Testing OpenAI’s o1, a model trained to reason step by step, the authors of “Embers of Autoregression” found that it substantially outperformed earlier models “in many cases” but “still displays the same qualitative trends”: its accuracy at decoding a shift cipher ranged from 47% in the lowest-probability case to 92% in the highest, and it used more reasoning tokens on low-probability examples [Claim LD01-SL-C37, version v1.0]. SocioLens. https://sociolens-web.pages.dev/claims/ld01-sl-c37/v1.0/

@misc{sociolens_ld01_sl_c37_v1_0,
  author = {{SocioLens project}},
  title = {Testing OpenAI’s o1, a model trained to reason step by step, the authors of “Embers of Autoregression” found that it substantially outperformed earlier models “in many cases” but “still displays the same qualitative trends”: its accuracy at decoding a shift cipher ranged from 47\% in the lowest-probability case to 92\% in the highest, and it used more reasoning tokens on low-probability examples.},
  year = {2026},
  date = {2026-10-16},
  howpublished = {\url{https://sociolens-web.pages.dev/claims/ld01-sl-c37/v1.0/}},
  note = {Claim LD01-SL-C37, version v1.0}
}

The link points to this version, whose record never changes. For a finding, cite the original study as well: SocioLens summarizes research and does not replace it.

Research-of-record provenance

Materialized by sociolens-materializer 0.6.0 at 2026-09-28T20:00:00.000Z; normalization contract SocioLens — Claim Value Vocabulary v1.0.2 — Pre-Spike Materialization Contract (v1.0.2). Content seal SHA-256 dc86dfb7d23915cdf6452778d3e6a3963abb0f1a8a13ad075f5a46900749499d.

Repository-authored inputs (not Drive exports)

SocioLens claim — LD01-SL-C37 v1.0 (v1.0) — written in the SocioLens repository by Project Lead (Claude, AI) on 2026-10-16; no Drive identity exists. Authority: AGENTS.md "Authority" (the Project Lead decides research content within the research-integrity rules) and the Owner's choice in the SocioLens project chat on 16 October 2026 to build the LD-01 debate; a new Claim, first version, written from the project's LD-01 source notes, checked against full texts, official web pages or official abstracts. Block “ENTRY 37 — LD01-SL-C37” SHA-256 64bf3527da5d3b35ce70be5c531fb4288ff9d41d4835f787706ff0a0099d59fe; fixture fixtures/authored/claims/LD01-SL-C37--v1.0.

Builder decisions (not Drive data)

VERSION = v1.0 — basis: Claim version taken from the repository-authored claim fixture version (SocioLens claim — LD01-SL-C37 v1.0); the first version of a new Claim of the LD-01 debate
CURRENT_POINTER = true — basis: repository-authored claim "SocioLens claim — LD01-SL-C37 v1.0" (Project Lead (Claude, AI), 2026-10-16); basis: AGENTS.md "Authority" (the Project Lead decides research content within the research-integrity rules) and the Owner's choice in the SocioLens project chat on 16 October 2026 to build the LD-01 debate; a new Claim, first version, written from the project's LD-01 source notes, checked against full texts, official web pages or official abstracts
FIELD_6_SOURCE_REFERENCE = SRC-SL-MCCOY-ET-AL-O1-ARXIV-2410-01792--v1.0 — basis: field 6 of ENTRY 37 — LD01-SL-C37 cites "McCoy, Yao, Friedman, Hardy & Griffiths, “When a language model is optimized for reasoning, does it still show embers of autoregression? An analysis of OpenAI o1,” arXiv:2410.01792, 2024"; the only field-6 segment of this ClaimVersion's repository-authored ledger block.
SOURCE_PROVENANCE_PIN = SRC-SL-MCCOY-ET-AL-O1-ARXIV-2410-01792--v1.0 — basis: explicit version-pinned public provenance of this ClaimVersion (Amendment v1.2 §6): the pinned field-6 Source record "McCoy, Yao, Friedman, Hardy & Griffiths, “When a language model is optimized for reasoning, does it still show embers of autoregression? An analysis of OpenAI o1,” arXiv:2410.01792, 2024".