Papers › Robust Adaptation of Large Multimodal Models for Retrieval Augmented Hateful Meme Detection

Robust Adaptation of Large Multimodal Models for Retrieval Augmented Hateful Meme Detection

18 Feb 2025arXiv:2502.13061archive 2025-07-28

Jingbiao Mei, Jinghong Chen, Guangyu Yang, Weizhe Lin, Bill Byrne

Hateful memes have become a significant concern on the Internet, necessitating robust automated detection systems. While LMMs have shown promise in hateful meme detection, they face notable challenges like sub-optimal performance and limited out-of-domain generalization capabilities. Recent studies further reveal the limitations of both SFT and in-context learning when applied to LMMs in this setting. To address these issues, we propose a robust adaptation framework for hateful meme detection that enhances in-domain accuracy and cross-domain generalization while preserving the general vision-language capabilities of LMMs. Experiments on six meme classification datasets show that our approach achieves state-of-the-art performance, outperforming larger agentic systems. Moreover, our method generates higher-quality rationales for explaining hateful content compared to standard SFT, enhancing model interpretability.

PaperPDFConference PDFCodeCode Syntology ran

In Syntology View this paper on Syntology: its repositories, every harvested function with whether it ran, its licence and the call to fetch it.

Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

For agents, Syntology's MCP tool lists every function and class Syntology harvested from this paper and whether it ran (how to connect): get_harvested_code_for_paper(arxiv_id="2502.13061")

Code

Syntology Ran 2 of 4 code samples harvested from 1 repository linked to this paper; 2 have no recorded run. Of those that ran: 1 ran · our draft was wrong; 1 ran · fixture could not drive it.

By repository: official repository: 4 samples from 1 repository, 2 ran. The run record, sample by sample. “Ran” means executed on a synthesized input, not that the code is correct or reproduces the paper.

JingbiaoMei/RGCL officialmentioned on GitHubpytorch report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

4 samples harvested; 2 ran; 0 honoured the contract we drafted; 2 have no recorded run. Read from Syntology's graph 2026-09-25; that is when this build read the record, not when the samples ran.

1ran · our draft was wrong
1ran · fixture could not drive it
2unverified

Licence: 0 of the 4 samples are pointer only, meaning Syntology does not serve that copy's text. This page shows no code text for any sample; each one links to its file in the repository.

Harvested from JingbiaoMei/RGCL. “Ran” means the sample executed on a synthesized input. It does not mean the output is correct, and nothing here reproduces the paper's results. “Honoured” and “violated” refer to a contract Syntology drafted from the code itself; “our draft was wrong” and “fixture could not drive it” are failures of Syntology's instrument, not of the code.

Each sample ends with its code_sha256, Syntology's identity for that exact code. An agent fetches the stored sample with Syntology's MCP tool get_code(code_sha256="…") (how to connect); click an identity to copy that call.

Repository labels, per sample. official repository: The archive marks this repository official for the paper. named in the paper: The archive records that the paper mentions this repository; it is not marked official. community (archive-listed): In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper. found in paper text by Syntology: Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted. community: Not in the archive's code links for this paper; a community repository Syntology harvested. Samples from a repository marked official are listed first. Licence labels name the repository's licence as recorded at harvest. “Pointer only” means Syntology does not serve that copy's text, for one of four reasons: no licence file was found; the licence was not identified; the licence is recorded as permissive but that copy's record is not marked cleared; or the licence is outside the permissive list Syntology serves text under (MIT, Apache-2.0, BSD and similar). Some licences outside that list permit redistribution, such as WTFPL, and GPL-3.0 under its conditions; they are simply not on the list. Hover a licence label for the reason. File links open the file on GitHub at the default branch, which may have changed since the harvest.

compute_l2 JingbiaoMei/RGCL/src/model/loss.py official repository ran · fixture could not drive it fingerprinted Apache-2.0 (permissive) · b79d1b8058e01331 · report
sparse_retrieve_hard_negatives_pseudo_positive JingbiaoMei/RGCL/src/model/loss.py official repository ran · our draft was wrong Apache-2.0 (permissive) · a20178fd77f8b5a5 · report
compute_loss JingbiaoMei/RGCL/src/model/loss.py official repository unverified Apache-2.0 (permissive) · e29c3311f99491d1 · report
dense_retrieve_hard_negatives_pseudo_positive JingbiaoMei/RGCL/src/model/loss.py official repository unverified Apache-2.0 (permissive) · 20d79017dbff7d0c · report

Tasks

Contrastive LearningDomain GeneralizationHateful Meme ClassificationIn-Context LearningMeme ClassificationRetrieval

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Hateful Meme Classification HarMeme RA-HMD (Qwen2VL-7B) AUROC 93.2 #1 of 8 Archive leaderboard report
Hateful Meme Classification HarMeme RA-HMD (Qwen2VL-7B) Accuracy 88.1 #1 of 8 Archive leaderboard report
Hateful Meme Classification HarMeme RA-HMD (Qwen2VL-2B) AUROC 92.9 #2 of 8 Archive leaderboard report
Hateful Meme Classification HarMeme RA-HMD (Qwen2VL-2B) Accuracy 87.7 #2 of 8 Archive leaderboard report
Hateful Meme Classification Harm-P RA-HMD (Qwen2-VL-7B) Accuracy 91.6 #1 of 5 Archive leaderboard report
Hateful Meme Classification Harm-P RA-HMD (Qwen2-VL-7B) F1 91.1 #1 of 5 Archive leaderboard report
Hateful Meme Classification Hateful Memes RA-HMD (Qwen2-VL-7B) AUROC 91.1 #1 of 1 Archive leaderboard report
Hateful Meme Classification PrideMM RA-HMD (Qwen2-VL-7B) Accuracy 78.1 #1 of 7 Archive leaderboard report
Hateful Meme Classification PrideMM RA-HMD (Qwen2-VL-7B) F1 78.4 #1 of 7 Archive leaderboard report
Hateful Meme Classification PrideMM RA-HMD (Qwen2-VL-2B) Accuracy 76.0 #5 of 7 Archive leaderboard report
Hateful Meme Classification PrideMM RA-HMD (Qwen2-VL-2B) F1 76.7 #5 of 7 Archive leaderboard report
Meme Classification Hateful Memes RA-HMD (Qwen2-VL-7B) Accuracy 0.821 #1 of 17 Archive leaderboard report
Meme Classification Hateful Memes RA-HMD (Qwen2-VL-7B) ROC-AUC 0.911 #1 of 17 Archive leaderboard report
Meme Classification Hateful Memes RA-HMD (LLaVA-1.5-7B) Accuracy 0.809 #2 of 17 Archive leaderboard report
Meme Classification Hateful Memes RA-HMD (LLaVA-1.5-7B) ROC-AUC 0.897 #2 of 17 Archive leaderboard report
Meme Classification Hateful Memes RA-HMD (Qwen2-VL-2B) Accuracy 0.791 #4 of 17 Archive leaderboard report
Meme Classification Hateful Memes RA-HMD (Qwen2-VL-2B) ROC-AUC 0.884 #4 of 17 Archive leaderboard report
Meme Classification MultiOFF RA-HMD (Qwen2-VL-7B) Accuracy 71.1 #1 of 5 Archive leaderboard report
Meme Classification MultiOFF RA-HMD (Qwen2-VL-7B) F1 64.8 #1 of 5 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Methods

Contrastive LearningSFT

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections