Adding a clause that seems relevant but does not change the answer to grade-school maths problems cut accuracy by up to 65 percentage points across 25 models; GPT-4o scored 95.2% on the full GSM8K test set and 63.1% with the clause added, and OpenAI’s o1-preview 94.9% and 77.4%.

SocioLens project,

Claim LD01-SL-C20 (v1.0)

Provenance and status

Claim type: Estimate
Publication status: Eligible for publication, with a qualifier
Verification state: Checked against the primary source. 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 finding that models cannot reason; the authors state their pattern-matching reading as a hypothesis
Limitations: Read in arXiv version 2 (27 August 2025); no human baseline on the irrelevant clauses; 2024 models; the authors work for a company that sells AI products; a 2026 re-analysis of 20 open-weight models (Długosz, Oliveira & Díaz-Rodríguez, arXiv:2605.28700) disputes the statistical basis of the finding on changed names and numbers but leaves the irrelevant-clause variant out of scope
Source (as cited): Mirzadeh, Alizadeh, Shahrokhi, Tuzel, Bengio & Farajtabar, “GSM-Symbolic: Understanding the Limitations of Mathematical Reasoning in Large Language Models,” ICLR 2025
Source type: Peer-reviewed conference paper (recorded as “Peer-reviewed conference paper”)
Institutional position: Industry research (Apple; one author at Washington State University)
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

Mirzadeh, Alizadeh, Shahrokhi, Tuzel, Bengio & Farajtabar, “GSM-Symbolic: Understanding the Limitations of Mathematical Reasoning in Large Language Models,” ICLR 2025 (Peer-reviewed conference paper)
Read the original

Version

This page shows version v1.0. Its record never changes, even when newer evidence arrives; the site may add a note to the page or withhold a figure not cleared for public use.

This pinned version is also the current version.

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 (QUALIFIED) (as printed: “ELIGIBLE WITH QUALIFIER”)
Verification state (coded): CONTENT_VERIFIED_PRIMARY
Source type (coded): PEER_REVIEWED_CONFERENCE_PAPER
Maturity (coded): EMERGING

#	Canonical field	Canonical value (verbatim)	Normalized (Claim Value Vocabulary v1.0.2)
1	CLAIM ID	LD01-SL-C20	—
2	PUBLIC-FACING CLAIM	Adding a clause that seems relevant but does not change the answer to grade-school maths problems cut accuracy by up to 65 percentage points across 25 models; GPT-4o scored 95.2% on the full GSM8K test set and 63.1% with the clause added, and OpenAI’s o1-preview 94.9% and 77.4%.	—
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	Mirzadeh, Alizadeh, Shahrokhi, Tuzel, Bengio & Farajtabar, “GSM-Symbolic: Understanding the Limitations of Mathematical Reasoning in Large Language Models,” ICLR 2025	source_ids: SRC-SL-MIRZADEH-ET-AL-GSM-SYMBOLIC-ARXIV-2410-05229
7	SOURCE TYPE	Peer-reviewed conference paper	classes: PEER_REVIEWED_CONFERENCE_PAPER
8	INSTITUTIONAL POSITION	Industry research (Apple; one author at Washington State University)	—
9	SCOPE	Grade-school maths word problems in English; models available in 2024	—
10	DESIGN	Templates that vary names and numbers in benchmark problems; harder variants; an irrelevant-clause variant	—
11	SAMPLE / DATASET	Templates built from 100 GSM8K problems; 25 models	—
12	MODEL / TOOL / VERSION	25 open and proprietary models, including GPT-4o and o1-preview	—
13	COMPARATOR	The full GSM8K test set and the 100 template problems without the added clause (95.0% for GPT-4o, 96.0% for o1-preview)	—
14	MAIN RESULT	Accuracy varies across versions of the same problem; drops of up to 65 percentage points with an irrelevant clause; GPT-4o 95.2% (full GSM8K) and 63.1% (with the clause); o1-preview 94.9% and 77.4%	—
15	INFERENTIAL TARGET	Whether models’ mathematical reasoning is robust to changes that do not alter the problem	—
16	DOES NOT LICENSE	A finding that models cannot reason; the authors state their pattern-matching reading as a hypothesis	—
17	VERIFICATION STATE	CONTENT VERIFIED	state: CONTENT_VERIFIED_PRIMARY · verified_by: UNKNOWN
18	LIMITATIONS	Read in arXiv version 2 (27 August 2025); no human baseline on the irrelevant clauses; 2024 models; the authors work for a company that sells AI products; a 2026 re-analysis of 20 open-weight models (Długosz, Oliveira & Díaz-Rodríguez, arXiv:2605.28700) disputes the statistical basis of the finding on changed names and numbers but leaves the irrelevant-clause variant out of scope	—
19	CONTRADICTORY / SUPPORTING SOURCES	LD01-SL-C18, LD01-SL-C21	—
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?	Humans showing similar drops on the same items, or newer models showing none	—
24	PUBLICATION STATUS	ELIGIBLE WITH QUALIFIER	state: ELIGIBLE · qualifiers: QUALIFIED
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)

“Mirzadeh, Alizadeh, Shahrokhi, Tuzel, Bengio & Farajtabar, “GSM-Symbolic: Understanding the Limitations of Mathematical Reasoning in Large Language Models,” ICLR 2025” → Mirzadeh et al. (2025), “GSM-Symbolic: Understanding the…” (SRC-SL-MIRZADEH-ET-AL-GSM-SYMBOLIC-ARXIV-2410-05229--v1.0) — Mirzadeh, Alizadeh, Shahrokhi, Tuzel, Bengio & Farajtabar, “GSM-Symbolic: Understanding the Limitations of Mathematical Reasoning in Large Language Models,” ICLR 2025; class PEER_REVIEWED_CONFERENCE_PAPER (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). Adding a clause that seems relevant but does not change the answer to grade-school maths problems cut accuracy by up to 65 percentage points across 25 models; GPT-4o scored 95.2% on the full GSM8K test set and 63.1% with the clause added, and OpenAI’s o1-preview 94.9% and 77.4% [Claim LD01-SL-C20, version v1.0]. SocioLens. https://sociolens-web.pages.dev/claims/ld01-sl-c20/v1.0/

@misc{sociolens_ld01_sl_c20_v1_0,
  author = {{SocioLens project}},
  title = {Adding a clause that seems relevant but does not change the answer to grade-school maths problems cut accuracy by up to 65 percentage points across 25 models; GPT-4o scored 95.2\% on the full GSM8K test set and 63.1\% with the clause added, and OpenAI’s o1-preview 94.9\% and 77.4\%.},
  year = {2026},
  date = {2026-10-16},
  howpublished = {\url{https://sociolens-web.pages.dev/claims/ld01-sl-c20/v1.0/}},
  note = {Claim LD01-SL-C20, 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 ad68423f795cfc05a5116a4a7266300f9b43b9adc5e70dfc800830143bc25b95.

Repository-authored inputs (not Drive exports)

SocioLens claim — LD01-SL-C20 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 20 — LD01-SL-C20” SHA-256 79c0870a6ab9c0e19ea468f0d8c1fc24faee378c8d70ed05f4c7e3660797b8f0; fixture fixtures/authored/claims/LD01-SL-C20--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-C20 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-C20 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-MIRZADEH-ET-AL-GSM-SYMBOLIC-ARXIV-2410-05229--v1.0 — basis: field 6 of ENTRY 20 — LD01-SL-C20 cites "Mirzadeh, Alizadeh, Shahrokhi, Tuzel, Bengio & Farajtabar, “GSM-Symbolic: Understanding the Limitations of Mathematical Reasoning in Large Language Models,” ICLR 2025"; the only field-6 segment of this ClaimVersion's repository-authored ledger block.
SOURCE_PROVENANCE_PIN = SRC-SL-MIRZADEH-ET-AL-GSM-SYMBOLIC-ARXIV-2410-05229--v1.0 — basis: explicit version-pinned public provenance of this ClaimVersion (Amendment v1.2 §6): the pinned field-6 Source record "Mirzadeh, Alizadeh, Shahrokhi, Tuzel, Bengio & Farajtabar, “GSM-Symbolic: Understanding the Limitations of Mathematical Reasoning in Large Language Models,” ICLR 2025".