# Do large language models scrutinise what they review? A multimodal audit of scoring calibration, error detection, and author-identity effects

arXiv: https://arxiv.org/abs/2608.28626 · HTML: https://mlsanity.com/paper/2608.28626/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 4 of its 11 references that are in the map.
- First posted on arXiv 31 Jul 2026.

## Builds on (most cited first, 4 of 4)

- [Can large language models provide useful feedback on research papers? A large-scale empirical analysis](https://mlsanity.com/paper/2310.01783/) (2023)
- [Monitoring AI-Modified Content at Scale: A Case Study on the Impact of ChatGPT on AI Conference Peer Reviews](https://mlsanity.com/paper/2403.07183/) (2024)
- [ReviewerGPT? An Exploratory Study on Using Large Language Models for Paper Reviewing](https://mlsanity.com/paper/2306.00622/) (2023)
- [Large language models for automated scholarly paper review: A survey](https://mlsanity.com/paper/2501.10326/) (2025)

## Abstract (arXiv)

Large language models (LLMs) are increasingly used to generate peer reviews, prompting examination of their capacity for critical evaluation. This study evaluates two multimodal LLMs, Qwen2.5-VL-72B and Pixtral-Large-124B, as reviewers across 165 submissions to the 2026 International Conference on Learning Representations, a venue that postdates both models' training cutoffs. Manuscripts were presented to both models with author identities blinded, replaced with high-prestige affiliations, or replaced with low-prestige affiliations, and in either text-only or text-with-figure format. Additionally, 145 verifiably detectable errors were inserted into 55 manuscripts to assess error identification under natural and verification-oriented prompts. Across all manuscript groups, including rejected submissions, LLM scores ranged from 7.0 to 8.1, whereas human mean scores ranged from 3.4 to 6.8. The models detected 12.1% of the verified errors under natural prompting, and a one-sentence verification instruction increased detection to 22.2%; however, 78% of the errors remained undetected. Providing figures reduced error detection while increasing review scores. No visual error was reliably verified against its corresponding figure, and half of the text-only reviews described figures that were not provided. Author identity did not influence either review scores or error detection. LLM editorial decisions exactly matched those produced by simple score averaging.