@inproceedings{li-etal-2026-m2note,
title = "{M}$^2${N}ote: Continual Evolution of Vision Language Models via Mistake Notebook Learning",
author = "Li, Haiwen and
Tang, Jing and
Chen, Rui and
Sun, Lei and
Chen, Zining and
Chu, Xiangxiang",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-main.716/",
pages = "16611--16629",
ISBN = "979-8-89176-459-0",
abstract = "Vision Language Models (VLMs) have demonstrated remarkable capabilities in multimodal reasoning tasks, yet they still suffer from recurring failures, such as skipping key visual checks, misapplying domain rules, and hallucinating unsupported concepts. Most existing solutions rely on supervised fine-tuning (SFT) and reinforcement learning (RL), which are expensive to iterate and can be brittle under distribution shift. To this end, we propose $\textbf{M}$ultimodal $\textbf{M}$istake $\textbf{Note}$book Learning (M$^2$Note), a training-free continual evolution framework that externalizes learning into an editable memory. M$^2$Note transforms failed trajectories into compact subject{--}guidance notes: the subject summarizes the underlying domain and concept, while the guidance provides actionable verification steps that can be reused in future inference. At test time, M$^2$Note retrieves relevant notes via multimodal retrieval-augmented generation (RAG) and appends them to the model context, steering reasoning away from previously observed pitfalls. To stabilize continual evolution, we adopt batch-level post-verification with rollback, which commits notebook edits only if they improve performance on the same batch, reducing noisy updates and preventing regressions. M$^2$Note supports both $\textit{self-evolving}$, where the same VLM acts as solver and supervisor, and $\textit{cross-model evolving}$, where a stronger supervisor guides a weaker solver, enabling capability transfer without weight updates. Experiments on six multimodal reasoning benchmarks show consistent improvements across domains and backbones, while achieving strong cost and sample efficiency and remaining complementary to Chain-of-Thought (CoT) prompting. The code is available at https://github.com/AMAP-ML/M2Note."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="li-etal-2026-m2note">
<titleInfo>
<title>M²Note: Continual Evolution of Vision Language Models via Mistake Notebook Learning</title>
</titleInfo>
<name type="personal">
<namePart type="given">Haiwen</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jing</namePart>
<namePart type="family">Tang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Rui</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Lei</namePart>
<namePart type="family">Sun</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zining</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xiangxiang</namePart>
<namePart type="family">Chu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hung-Yi</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sunipa</namePart>
<namePart type="family">Dev</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jessy</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Desmond</namePart>
<namePart type="family">Elliott</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-459-0</identifier>
</relatedItem>
<abstract>Vision Language Models (VLMs) have demonstrated remarkable capabilities in multimodal reasoning tasks, yet they still suffer from recurring failures, such as skipping key visual checks, misapplying domain rules, and hallucinating unsupported concepts. Most existing solutions rely on supervised fine-tuning (SFT) and reinforcement learning (RL), which are expensive to iterate and can be brittle under distribution shift. To this end, we propose Multimodal Mistake Notebook Learning (M²Note), a training-free continual evolution framework that externalizes learning into an editable memory. M²Note transforms failed trajectories into compact subject–guidance notes: the subject summarizes the underlying domain and concept, while the guidance provides actionable verification steps that can be reused in future inference. At test time, M²Note retrieves relevant notes via multimodal retrieval-augmented generation (RAG) and appends them to the model context, steering reasoning away from previously observed pitfalls. To stabilize continual evolution, we adopt batch-level post-verification with rollback, which commits notebook edits only if they improve performance on the same batch, reducing noisy updates and preventing regressions. M²Note supports both self-evolving, where the same VLM acts as solver and supervisor, and cross-model evolving, where a stronger supervisor guides a weaker solver, enabling capability transfer without weight updates. Experiments on six multimodal reasoning benchmarks show consistent improvements across domains and backbones, while achieving strong cost and sample efficiency and remaining complementary to Chain-of-Thought (CoT) prompting. The code is available at https://github.com/AMAP-ML/M2Note.</abstract>
<identifier type="citekey">li-etal-2026-m2note</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.716/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>16611</start>
<end>16629</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T M²Note: Continual Evolution of Vision Language Models via Mistake Notebook Learning
%A Li, Haiwen
%A Tang, Jing
%A Chen, Rui
%A Sun, Lei
%A Chen, Zining
%A Chu, Xiangxiang
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-459-0
%F li-etal-2026-m2note
%X Vision Language Models (VLMs) have demonstrated remarkable capabilities in multimodal reasoning tasks, yet they still suffer from recurring failures, such as skipping key visual checks, misapplying domain rules, and hallucinating unsupported concepts. Most existing solutions rely on supervised fine-tuning (SFT) and reinforcement learning (RL), which are expensive to iterate and can be brittle under distribution shift. To this end, we propose Multimodal Mistake Notebook Learning (M²Note), a training-free continual evolution framework that externalizes learning into an editable memory. M²Note transforms failed trajectories into compact subject–guidance notes: the subject summarizes the underlying domain and concept, while the guidance provides actionable verification steps that can be reused in future inference. At test time, M²Note retrieves relevant notes via multimodal retrieval-augmented generation (RAG) and appends them to the model context, steering reasoning away from previously observed pitfalls. To stabilize continual evolution, we adopt batch-level post-verification with rollback, which commits notebook edits only if they improve performance on the same batch, reducing noisy updates and preventing regressions. M²Note supports both self-evolving, where the same VLM acts as solver and supervisor, and cross-model evolving, where a stronger supervisor guides a weaker solver, enabling capability transfer without weight updates. Experiments on six multimodal reasoning benchmarks show consistent improvements across domains and backbones, while achieving strong cost and sample efficiency and remaining complementary to Chain-of-Thought (CoT) prompting. The code is available at https://github.com/AMAP-ML/M2Note.
%U https://aclanthology.org/2026.emnlp-main.716/
%P 16611-16629
Markdown (Informal)
[M2Note: Continual Evolution of Vision Language Models via Mistake Notebook Learning](https://aclanthology.org/2026.emnlp-main.716/) (Li et al., EMNLP 2026)
ACL