Off-the-shelf ChatGPT and GPT-4 reached only 55% and 59% accuracy on 3-digit by 3-digit multiplication, and accuracy on compositional tasks fell towards zero as complexity grew; the authors hypothesize that transformers reduce multi-step reasoning to “linearized path matching.”

SocioLens project,

Claim LD01-SL-C21 (v1.0)

Provenance and status

Claim type: Estimate and Interpretation
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 claim about later reasoning models
Limitations: Read in arXiv version 3 (31 October 2023); 2023 models without later reasoning methods; the pattern-matching reading is inferred from frequency correlations
Source (as cited): Dziri et al., “Faith and Fate: Limits of Transformers on Compositionality,” NeurIPS 2023
Source type: Peer-reviewed conference paper (recorded as “Peer-reviewed conference paper”)
Institutional position: Independent academic and nonprofit research (Allen Institute for AI, University of Washington, University of Southern California, University of Chicago)
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

Dziri et al., “Faith and Fate: Limits of Transformers on Compositionality,” NeurIPS 2023 (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 + INTERPRETATION (as printed: “ESTIMATE / INTERPRETATION”)
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-C21	—
2	PUBLIC-FACING CLAIM	Off-the-shelf ChatGPT and GPT-4 reached only 55% and 59% accuracy on 3-digit by 3-digit multiplication, and accuracy on compositional tasks fell towards zero as complexity grew; the authors hypothesize that transformers reduce multi-step reasoning to “linearized path matching.”	—
3	CLAIM TYPE	ESTIMATE / INTERPRETATION	base_types: ESTIMATE, INTERPRETATION
4	RESEARCH STRAND	B — Evidence that could tell the accounts apart	—
5	MATURITY	EMERGING	states: EMERGING
6	SOURCE	Dziri et al., “Faith and Fate: Limits of Transformers on Compositionality,” NeurIPS 2023	source_ids: SRC-SL-DZIRI-ET-AL-FAITH-FATE-ARXIV-2305-18654
7	SOURCE TYPE	Peer-reviewed conference paper	classes: PEER_REVIEWED_CONFERENCE_PAPER
8	INSTITUTIONAL POSITION	Independent academic and nonprofit research (Allen Institute for AI, University of Washington, University of Southern California, University of Chicago)	—
9	SCOPE	Three compositional tasks (multiplication, logic grid puzzles, a dynamic-programming problem); 2023 models	—
10	DESIGN	Experiments with zero-shot, few-shot, scratchpad and fine-tuned models, and a theoretical argument on error propagation	—
11	SAMPLE / DATASET	Three task families at increasing complexity	—
12	MODEL / TOOL / VERSION	GPT-3 (fine-tuned), ChatGPT, GPT-4	—
13	COMPARATOR	Task complexity; in-distribution and out-of-distribution problems	—
14	MAIN RESULT	55% (ChatGPT) and 59% (GPT-4) on 3×3-digit multiplication; near-perfect in-distribution after fine-tuning but failure out of distribution	—
15	INFERENTIAL TARGET	Whether transformers solve compositional tasks systematically	—
16	DOES NOT LICENSE	A claim about later reasoning models	—
17	VERIFICATION STATE	CONTENT VERIFIED	state: CONTENT_VERIFIED_PRIMARY · verified_by: UNKNOWN
18	LIMITATIONS	Read in arXiv version 3 (31 October 2023); 2023 models without later reasoning methods; the pattern-matching reading is inferred from frequency correlations	—
19	CONTRADICTORY / SUPPORTING SOURCES	LD01-SL-C20, LD01-SL-C22	—
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?	Models that keep accuracy as the size of compositional problems grows	—
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)

“Dziri et al., “Faith and Fate: Limits of Transformers on Compositionality,” NeurIPS 2023” → Dziri et al. (2023), “Faith and Fate: Limits of…” (SRC-SL-DZIRI-ET-AL-FAITH-FATE-ARXIV-2305-18654--v1.0) — Dziri et al., “Faith and Fate: Limits of Transformers on Compositionality,” NeurIPS 2023; 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). Off-the-shelf ChatGPT and GPT-4 reached only 55% and 59% accuracy on 3-digit by 3-digit multiplication, and accuracy on compositional tasks fell towards zero as complexity grew; the authors hypothesize that transformers reduce multi-step reasoning to “linearized path matching.” [Claim LD01-SL-C21, version v1.0]. SocioLens. https://sociolens-web.pages.dev/claims/ld01-sl-c21/v1.0/

@misc{sociolens_ld01_sl_c21_v1_0,
  author = {{SocioLens project}},
  title = {Off-the-shelf ChatGPT and GPT-4 reached only 55\% and 59\% accuracy on 3-digit by 3-digit multiplication, and accuracy on compositional tasks fell towards zero as complexity grew; the authors hypothesize that transformers reduce multi-step reasoning to “linearized path matching.”},
  year = {2026},
  date = {2026-10-16},
  howpublished = {\url{https://sociolens-web.pages.dev/claims/ld01-sl-c21/v1.0/}},
  note = {Claim LD01-SL-C21, 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 a66eb961ff8e9dff08a5ae5925f00110418b1bb1eb4127592b6769d29f460788.

Repository-authored inputs (not Drive exports)

SocioLens claim — LD01-SL-C21 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 21 — LD01-SL-C21” SHA-256 93d619f6555d8e3ff954f31982f70cb9e9d8bc090ba7c595ab2915083295234f; fixture fixtures/authored/claims/LD01-SL-C21--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-C21 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-C21 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-DZIRI-ET-AL-FAITH-FATE-ARXIV-2305-18654--v1.0 — basis: field 6 of ENTRY 21 — LD01-SL-C21 cites "Dziri et al., “Faith and Fate: Limits of Transformers on Compositionality,” NeurIPS 2023"; the only field-6 segment of this ClaimVersion's repository-authored ledger block.
SOURCE_PROVENANCE_PIN = SRC-SL-DZIRI-ET-AL-FAITH-FATE-ARXIV-2305-18654--v1.0 — basis: explicit version-pinned public provenance of this ClaimVersion (Amendment v1.2 §6): the pinned field-6 Source record "Dziri et al., “Faith and Fate: Limits of Transformers on Compositionality,” NeurIPS 2023".