Cross-Context Review: Improving LLM Output Quality by Separating Production and Review Sessions

Tae-Eun Song

cs.CLMar 12, 2026

Cross-Context Review: Improving LLM Output Quality by Separating Production and Review Sessions

Authors: Tae-Eun Song

Organizations: Daejeon Jungang Cheonggua Co., Ltd.

Abstract

Large language models struggle to catch errors in their own outputs when the review happens in the same session that produced them. This paper introduces Cross-Context Review (CCR), a straightforward method where the review is conducted in a fresh session with no access to the production conversation history. We ran a controlled experiment: 30 artifacts (code, technical documents, presentation scripts) with 150 injected errors, tested under four review conditions -- same-session Self-Review (SR), repeated Self-Review (SR2), context-aware Subagent Review (SA), and Cross-Context Review (CCR). The central result is that a second review helps only when it happens in a fresh session: CCR (F1 28.6%) outperforms a second review in the same session (SR2, 21.7%) robustly, both in the first run (paired t, p<0.001) and in the three-run average (Holm-adjusted p=0.004). This version updates the broader comparisons. Averaged across runs, and excluding one SR run whose records could not be verified, CCR is not significantly ahead of context-aware subagent review (SA, 23.8%; p=0.057) or of a single same-session review (SR, 27.1%; p=0.26); the first version's advantages over these two baselines came from run 1. CCR needs no infrastructure and costs one extra session.

Figures & tables

Condition	Abbr.	Context	Purpose
Self-Review	SR	Full history	Baseline
Self-Review $\times$ 2	SR2	Full + 1st review	Repetition ctrl
Subagent Review	SA	Prompt + artifact	Context ctrl
Cross-Context	CCR	Artifact only	Our method

Table 1: Review conditions and their available context.

Cond.	Finds	TP	FP	FN	Prec.	Rec.	F1
CCR	4.5	1.4	3.1	3.6	31.5	27.1	28.6
SR †	5.6	1.4	4.2	3.6	26.2	28.3	27.1
SA	4.3	1.1	3.2	3.9	27.4	21.8	23.8
SR2	5.5	1.1	4.4	3.9	21.0	22.7	21.7

Table 2: Overall results (

N

=90 per condition; † SR uses runs 1–2 only,

N

=60, Section 6 ; v1 values with run 3: F1 24.5, Prec. 25.8, Rec. 24.2). Finds/TP/FP/FN are per-artifact averages; Prec./Rec./F1 in %.

Comparison	$t$	$p$	Cohen’s $d$	Sig.
SR vs CCR	$-$ 2.849	0.008	0.52	Yes**
SR2 vs CCR	$-$ 3.917	<0.001	0.72	Yes***
SA vs CCR	$-$ 3.119	0.004	0.57	Yes**
SR vs SR2	1.665	0.107	$-$ 0.30	No

Table 3: Statistical comparisons (paired

t

-test, Run 1).

t

is computed as the first minus the second condition in each row; Cohen’s

d

d_{z}

, the mean paired difference (second minus first) divided by its standard deviation.

Condition	Critical	Major	Minor
CCR	40%	29%	18%
SR †	32%	32%	22%
SA	30%	23%	16%
SR2	31%	26%	16%

Table 4: Detection rate by severity ( † SR runs 1–2 only; v1: 29/27/19%).

Category	CCR	SR	SA	SR2
Code	40.7%	37.0%	34.1%	31.9%
Document	24.5%	22.3%	17.4%	19.2%
Script	20.7%	22.1%	19.9%	14.0%

Table 5: F1 by artifact category (3-run average; SR runs 1–2 only; v1 SR: 36.0/18.2/19.5%).

Cond.	FACT	CONS	CTXT	RCVR	MISS
CCR	48%	37%	13%	19%	19%
SR †	50%	35%	18%	18%	20%
SA	41%	31%	9%	14%	13%
SR2	40%	31%	16%	12%	14%

Table 6: Detection rate by error type (

N

=90 per cell; † SR

N

=60, runs 1–2; v1 SR: 44/32/12/16/17%).

Figure 1: F1 scores by condition and artifact category (3-run average; SR runs 1–2 only). CCR leads on code and documents; SR leads on scripts.

Figure 2: Detection rate by error severity (SR runs 1–2 only). CCR is ahead of SR only on Critical errors (+8pp); SR is ahead on Major and Minor errors. Descriptive, not tested.

Cond.	Run 1	Run 2	Run 3
CCR	37.2	25.0	23.7
SR	25.7	28.6	(19.4, excluded)
SA	23.8	20.8	26.9
SR2	18.7	22.5	23.9

Table 7: F1 (%) per run, averaged over 30 artifacts.

Appendix figures & tables2 assets

Supplementary material from the paper’s appendix.

Appendix

Type	Code	Description
Factual	FACT	Wrong numbers/claims
Consistency	CONS	Internal contradictions
Contextual	CTXT	Environment issues
Audience	RCVR	Reader misinterpretation
Completeness	MISS	Important omission

Appendix

Table 8: Error taxonomy used in ground-truth annotation.

Round	Reviewer	Errors	Types
CCR-1	T2	15	Numerical, sections
CCR-2	T4	10	Authors, refs

Appendix

Table 9: Meta-validation: recursive CCR on this paper.

Explore similar work

When Does a Second Model Help? Cross-Model Review in LLM Verification
Oct 1, 2026Tae-Eun SongPeer Review Provenance
Daejeon Jungang Cheonggua Co., Ltd.
Cross-Model LLM Code Review: Should you use Claude to review Codex or vice versa?
Jul 22, 2026Zuodong Xiang, Yike Zhang, YueMing Zhang +1Code Quality Llm-As-A-Judge
University of California, Davis USA · Johns Hopkins University USA · California State University, Long Beach USA
CRJudgeBench: Can AI Detect Plausible but Invalid Code Reviews?
Sep 29, 2026Yue Pan, Jiawei Li, Ziyuan Zhang +2Code Quality Raw Judge Outputs
University College London · Amazon · Zhejiang University

Oct 1, 2026, cs.CLJ/K move · Enter open · S save

When Does a Second Model Help? Cross-Model Review in LLM Verification

Tae-Eun Song

Daejeon Jungang Cheonggua Co., Ltd.

Peer Review Provenance

Open paper AskPDF arXiv

Large language models now generate code, documentation, and analyses, and are increasingly used to review such output. We ask when a second review by a different model helps. Building on the author's earlier preprints, which varied context, repetition, and role structure within one model, we test model independence in a controlled experiment: 30 artifacts with 150 planted errors, 10 review conditions, and 900 review sessions with three reviewer models from two developers. In this experiment, (1) a top-tier cross-model reviewer is not significantly different in F1 from same-model review in a fresh session (CCR), which does not establish equivalence; (2) the two find partly different errors (Jaccard 41.2%); and (3) at two review calls, one CCR plus one cross-model review matches more planted errors than two CCR reviews (56.7% vs. 42.7%; Holm-adjusted p=.006), but not significantly more than two reviews by the top-tier cross-model reviewer, so model difference and reviewer capability are not separated. A lightweight cross-model reviewer scores no higher than same-model review. Withholding requirements from the reviewer raises F1 for the two lower tiers but not the top tier, in untested point estimates whose pattern depends on how failed sessions are scored. Before analysis we audited all session records, excluding one baseline run of uncertain provenance and 14 failed calls; results with all sessions are also reported. A partial check on public detector outputs from another benchmark neither replicates nor contradicts the main comparison. Records, artifacts, and scripts are available from the author on request.

Abstract

Explore similar work

When Does a Second Model Help? Cross-Model Review in LLM Verification

Cross-Model LLM Code Review: Should you use Claude to review Codex or vice versa?

CRJudgeBench: Can AI Detect Plausible but Invalid Code Reviews?