Claim LD06-SL-C09 (v1.0)

Meaning-preserving changes to prompt formatting changed few-shot accuracy by up to 76 percentage points for LLaMA-2-13B, and which formats work best correlated only weakly between models, which the authors say calls into question comparing models on one fixed prompt format.

Provenance and status

Claim type: ESTIMATE (as printed: “ESTIMATE”)
Publication status: ELIGIBLE (QUALIFIED) (as printed: “ELIGIBLE WITH QUALIFIER”)
Verification state: CONTENT_VERIFIED_PRIMARY (ABSTRACT_ONLY) (as printed: “CONTENT VERIFIED ABSTRACT”; verified by: UNKNOWN)
Does not license: The size of the effect for current frontier models
Source (field 6, as printed): Sclar, Choi, Tsvetkov & Suhr, “Quantifying Language Models’ Sensitivity to Spurious Features in Prompt Design,” ICLR 2024, arXiv:2310.11324
Source type (field 7): PEER_REVIEWED_CONFERENCE_PAPER (as printed: “Peer-reviewed conference paper”)
Institutional position: Independent academic
Maturity (claim evidence / publication maturity): MATURE (as printed: “MATURE”)
Recorded date: 2026-10-15
Currentness / correction: This is the current version. No correction or concern relation is recorded on its Sources.

Version

Pinned version v1.0 (entry LD06-SL-C09--v1.0) — this page never floats to newer evidence.

This pinned version is also the current version.

Versions:v1.0 · current

Pinned Sources (field 6, version-specific)

“Sclar, Choi, Tsvetkov & Suhr, “Quantifying Language Models’ Sensitivity to Spurious Features in Prompt Design,” ICLR 2024, arXiv:2310.11324” → SRC-SL-SCLAR-CHOI-TSVETKOV-SUHR-ICLR-2024--v1.0 — Sclar, Choi, Tsvetkov & Suhr, “Quantifying Language Models’ Sensitivity to Spurious Features in Prompt Design,” ICLR 2024, arXiv:2310.11324; class PEER_REVIEWED_CONFERENCE_PAPER (explicit public provenance pin)

All 25 canonical Claim fields (raw values and normalized values)

#	Canonical field	Canonical value (verbatim)	Normalized (Claim Value Vocabulary v1.0.2)
1	CLAIM ID	LD06-SL-C09	—
2	PUBLIC-FACING CLAIM	Meaning-preserving changes to prompt formatting changed few-shot accuracy by up to 76 percentage points for LLaMA-2-13B, and which formats work best correlated only weakly between models, which the authors say calls into question comparing models on one fixed prompt format.	—
3	CLAIM TYPE	ESTIMATE	base_types: ESTIMATE
4	RESEARCH STRAND	A — Benchmark validity and measurement	—
5	MATURITY	MATURE	states: MATURE
6	SOURCE	Sclar, Choi, Tsvetkov & Suhr, “Quantifying Language Models’ Sensitivity to Spurious Features in Prompt Design,” ICLR 2024, arXiv:2310.11324	source_ids: SRC-SL-SCLAR-CHOI-TSVETKOV-SUHR-ICLR-2024
7	SOURCE TYPE	Peer-reviewed conference paper	classes: PEER_REVIEWED_CONFERENCE_PAPER
8	INSTITUTIONAL POSITION	Independent academic	—
9	SCOPE	Classification and generation tasks in few-shot settings	—
10	DESIGN	Systematic variation of prompt formats	—
11	SAMPLE / DATASET	Many sampled formats per task	—
12	MODEL / TOOL / VERSION	Several widely used open-source models, including LLaMA-2-13B	—
13	COMPARATOR	Alternative plausible formats for the same task	—
14	MAIN RESULT	Up to 76 accuracy points between formats; sensitivity persists with model size, more examples and instruction tuning; format performance weakly correlated between models	—
15	INFERENTIAL TARGET	The stability of a benchmark score under arbitrary evaluation choices	—
16	DOES NOT LICENSE	The size of the effect for current frontier models	—
17	VERIFICATION STATE	CONTENT VERIFIED ABSTRACT	state: CONTENT_VERIFIED_PRIMARY · qualifiers: ABSTRACT_ONLY · verified_by: UNKNOWN
18	LIMITATIONS	Read in abstract only; 2023 open models	—
19	CONTRADICTORY / SUPPORTING SOURCES	The authors propose reporting a range across formats, a measurement repair	—
20	GEOGRAPHY / LANGUAGE	Not checked / English	—
21	DISTRIBUTIONAL LEDGER	Not analysed in this claim	—
22	R01 / FIELD-MAP LINKS	LD-06, D4b	—
23	WHAT WOULD CHANGE THIS CLAIM?	Evidence that current models and harnesses are robust to format	—
24	PUBLICATION STATUS	ELIGIBLE WITH QUALIFIER	state: ELIGIBLE · qualifiers: QUALIFIED
25	RECORDED DATE	2026-10-15	—

Field 22 references resolved against Field Map v0.2: LD-06 (LIVING_DEBATE); D4b (SUB_AREA)

Cited by Living Debate versions

LD-06 v0.1

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 f9ed22de67e436633be09fa9e55082397fff08ed08998d2ff90d46c27dd993a6.

Repository-authored inputs (not Drive exports)

SocioLens claim — LD06-SL-C09 v1.0 (v1.0) — written in the SocioLens repository by Project Lead (Claude, AI) on 2026-10-15; 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 15 October 2026 to build the LD-06 debate; a new Claim, first version, written from the project's LD-06 source notes, checked against full texts or official abstracts. Block “ENTRY 09 — LD06-SL-C09” SHA-256 bae3f2ff749552d988a6ddd0556f89eb8a2a8bead276f02be7ffabae54855b4c; fixture fixtures/authored/claims/LD06-SL-C09--v1.0.

Builder decisions (not Drive data)

VERSION = v1.0 — basis: Claim version taken from the repository-authored claim fixture version (SocioLens claim — LD06-SL-C09 v1.0); the first version of a new Claim of the LD-06 debate
CURRENT_POINTER = true — basis: repository-authored claim "SocioLens claim — LD06-SL-C09 v1.0" (Project Lead (Claude, AI), 2026-10-15); 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 15 October 2026 to build the LD-06 debate; a new Claim, first version, written from the project's LD-06 source notes, checked against full texts or official abstracts
FIELD_6_SOURCE_REFERENCE = SRC-SL-SCLAR-CHOI-TSVETKOV-SUHR-ICLR-2024--v1.0 — basis: field 6 of ENTRY 09 — LD06-SL-C09 cites "Sclar, Choi, Tsvetkov & Suhr, “Quantifying Language Models’ Sensitivity to Spurious Features in Prompt Design,” ICLR 2024, arXiv:2310.11324"; the only field-6 segment of this ClaimVersion's repository-authored ledger block.
SOURCE_PROVENANCE_PIN = SRC-SL-SCLAR-CHOI-TSVETKOV-SUHR-ICLR-2024--v1.0 — basis: explicit version-pinned public provenance of this ClaimVersion (Amendment v1.2 §6): the pinned field-6 Source record "Sclar, Choi, Tsvetkov & Suhr, “Quantifying Language Models’ Sensitivity to Spurious Features in Prompt Design,” ICLR 2024, arXiv:2310.11324".