Pith. sign in

REVIEW 3 major objections 5 minor 23 references

A new benchmark, AHA-MEMES, offers the first large-scale fine-grained Arabic hateful-meme dataset and shows fine-tuned multimodal models lead binary detection while fine-grained hate types remain hard.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · deepseek-v4-flash

2026-08-01 08:07 UTC pith:DEYDPRBD

load-bearing objection Useful new Arabic meme benchmark with a real but fixable selection-bias problem: the gold set is pre-filtered by Gemma-3-12B and the authors never show what that filter missed. the 3 major comments →

arxiv 2607.27393 v1 pith:DEYDPRBD submitted 2026-07-29 cs.CL cs.AI

AHA-Memes: A Fine-Grained Multimodal Benchmark for Understanding Hate in Arabic Memes

Mohamed Bayan Kmainasi , Ali Ezzat Shahroor , Abul Hasnat , Md. Rafiul Biswas , Wajdi Zaghouani , Firoj Alam This is my paper
classification cs.CL cs.AI
keywords Arabic memeshate speech detectionmultimodal benchmarkfine-grained annotationvision-language modelsmulti-label classificationhateful meme taxonomyArabic NLP
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper's aim is to give Arabic hateful-meme research a proper evaluation resource: 5,000 human-annotated memes labeled for both binary hatefulness and fine-grained, multi-label attack strategies, plus a ~66,000-meme silver-labeled corpus. The authors argue this is the first large-scale Arabic benchmark of its kind, filling a gap left by English-centric and propaganda-focused resources. Their experiments establish baselines across five model families. The headline results: a fine-tuned open-weight vision-language model reaches 0.768 binary macro-F1, while the best fine-grained hate-type macro-F1 is only 0.340, and zero-shot models miss most hateful memes. If the benchmark holds up, it gives the community a standardized testbed for a culturally grounded moderation problem.

Core claim

The central discovery is the dataset itself and what it reveals. AHA-MEMES uses a hierarchical annotation scheme: each meme receives a binary is_hateful judgment, hateful memes are then labeled with one or more attack types (Mocking, Incitement, Dehumanization, Slurs, Contempt, Inferiority, Exclusion, Other), and non-hateful memes with Humor, Sarcasm, or Other. The 37% hate rate in the gold set and the long-tailed subtype distribution (Mocking dominates; Exclusion has 17 examples) make the benchmark realistically imbalanced. On it, embedded Arabic text carries more of the signal than images alone, late fusion and Arabic-specific encoders are strong baselines, fine-tuning an open VLM gives th

What carries the argument

The load-bearing mechanism is the conditional hierarchical taxonomy. Because fine-grained labels are assigned only after the binary gate, subtype labels are consistent with the binary decision by construction, and multi-label attack types capture memes that combine several strategies (e.g., Slurs + Dehumanization). Data construction pairs this with a two-stage pipeline: near-duplicate removal, OCR text extraction (memes without text are discarded), weak pre-selection of 5K memes for gold annotation, and LLM-generated silver labels for the remaining ~66K memes. The silver set is explicitly positioned as auxiliary, not as evaluation data.

Load-bearing premise

The 5,000 gold memes were chosen from a pool that a weak classifier (Gemma-3-12B) had pre-labeled, so the benchmark is only as representative of Arabic hateful memes as that pre-filter was.

What would settle it

Collect a fresh, unfiltered random sample of Arabic memes from the same platforms, annotate them with the same bilingual guidelines, and compare hate rate, subtype distribution, and model rankings to AHA-MEMES; material differences would show the pre-selection step biased the benchmark.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Researchers get a standardized, human-verified test split of 1,000 Arabic memes for comparing hate-detection systems.
  • Fine-grained hate-type classification remains open: at 0.340 macro-F1, even the best system is far from usable for automated moderation of attack strategies.
  • Zero-shot VLMs under-predict hate (best open zero-shot model recalls only 31.8% of hateful memes), so unadapted models are unsafe as moderators; fine-tuning roughly doubles hate recall.
  • Text is the dominant modality in Arabic memes, and Arabic-specific encoders outperform multilingual ones, so OCR quality and dialect-aware text models matter.
  • The ~66K silver-labeled corpus offers a large in-domain resource for weak supervision, though its machine-generated labels need denoising.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The gold set was sampled from a pool pre-filtered by Gemma-3-12B, so the benchmark's difficulty and model rankings may not transfer to uncurated Arabic memes; a random-sample gold set would test this.
  • The silver corpus has a 0.87% hate rate versus 37% in the gold set, so any semi-supervised method trained on both must handle a major prior shift.
  • Co-occurrence patterns (Mocking pairs with Dehumanization, Incitement, Contempt, Slurs) suggest a structured or hierarchy-aware model could beat the flat multi-label classifiers the paper benchmarks.
  • The annotation scheme also collects targeted protected categories; target-aware detection is listed as future work but not evaluated, and the released metadata would allow it.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper introduces AHA-MEMES, a benchmark for hateful Arabic memes consisting of 5K human-annotated memes with binary hate labels and fine-grained multi-label annotations (hate types and non-hate subtypes), plus about 66K silver-labeled auxiliary memes. The authors define a hierarchical taxonomy, report annotation agreement (Cohen's κ = 0.91 for binary, 0.75 for hate-type), and provide baselines across text encoders, image encoders, late-fusion models, open-weight VLMs, closed VLMs, and few-shot ICL settings. The main empirical findings are that fine-tuned Qwen3-VL-8B reaches 0.768 binary macro-F1, the best fine-grained hate-type macro-F1 is 0.340 (Gemini-2.5-pro), and fine-grained classification remains difficult, especially for rare labels.

Significance. If the resource is representative, AHA-MEMES fills a clear gap: there is no existing large-scale Arabic meme benchmark with fine-grained, multi-label hate annotations. The annotation scheme is carefully designed, with explicit boundary rules separating protected-group hate from generic offensiveness, and the reported inter-annotator agreement is strong. The release of the gold set, the silver set, guidelines, and evaluation scripts is a useful community contribution. The broad model comparison is also valuable as a baseline suite. The main reservation is external validity: because the 5K gold set is a conditional sample selected from a Gemma-3-12B pre-filtered pool, the benchmark's difficulty distribution and model rankings may not transfer to uncurated Arabic memes. This is a load-bearing issue for the benchmark's generalizability claim, though it does not undermine the value of the annotated data itself.

major comments (3)
  1. [§3.3, Appendix E, Table 13] The gold set is not a random sample of Arabic memes. The paper states that Gemma-3-12B assigned preliminary binary labels to 71K memes and that 5K were then selected for manual annotation (§3.3), but the selection rule is never specified. The resulting gold set is 37.0% hateful, while the silver set (Table 13) is only 0.87% hateful, indicating heavy oversampling of Gemma-positive items. Without a documented sampling rule and a Gemma-vs-human confusion matrix, the benchmark's label distribution and the reported model rankings may be biased. In particular, if Gemma misses implicit or visually coded hate — and §5.4 shows that even Gemini-2.5-pro makes such errors — those cases are systematically underrepresented. Please specify the exact selection algorithm, report Gemma-human agreement on the 5K, and re-annotate or at least measure a random sample of Gemma-negative memes to quantify the fa
  2. [§5, Table 3, §4.3] The fine-grained evaluation is statistically fragile. Test support for several hate types is extremely small: Exclusion has n=3, Inferiority n=32, Contempt n=50 (Table 2). All experiments use a single seed (seed 42) and the best configuration is selected on dev and evaluated once on test (§4.3). No confidence intervals are reported. The gaps between top fine-grained systems (Gemini-2.5-pro 0.340 vs fine-tuned Qwen3-VL-8B 0.334) are well within the likely noise of these rare-label estimates. Please report per-label F1 with support counts, bootstrap confidence intervals, and ideally multiple seeds. This is necessary to support the claims about model ordering and the overall difficulty of fine-grained hate-type classification.
  3. [Table 2, Table 3 caption, Appendix E Task 2 prompt] The label space is defined inconsistently. Table 2 lists 11 fine-grained labels, with separate 'Other (H)' and 'Other (NH)' rows, and its caption says Humor, Sarcasm, and Other are non-hateful subtypes while the rest are hateful attack types. However, Table 3's caption says fine-grained macro-F1 is 'averaged over all ten categories,' and the Appendix E Task 2 prompt uses a 10-label taxonomy with a single shared 'Other' label. Section 3.4 also omits 'Other' from the hateful-type list. This ambiguity directly affects the reported fine-grained metric and the model training/prompting setup. Please align the taxonomy definition, the annotation instructions, the distribution table, and the evaluation code, and clarify how many labels are actually used in the fine-grained task.
minor comments (5)
  1. [§3.5] The phrase 'triple-annotated gold subset' is undefined. Were all 5K memes triple-annotated, or only a subset? Please clarify, since this affects the interpretation of the agreement statistics and the multi-label distribution analysis.
  2. [§3.2] The near-duplicate distance threshold of 3.6 is stated without justification or sensitivity analysis. Reporting how the threshold was chosen and how the dataset size/duplicate rate changes with it would improve reproducibility.
  3. [§3.2, Ethics] The paper says 'every retained meme contains both visual and textual modalities' because memes with no detectable OCR text were discarded, but the Limitations section does not mention this exclusion. Image-only memes are a meaningful part of Arabic meme culture; please state this scope restriction explicitly in the Limitations.
  4. [Appendix D, Figure 13] In Table 11 and the surrounding text, the fine-grained zero-shot baseline is reported as 0.176, while Figure 3 says fine-grained performance 'improves modestly.' The numbers are consistent, but the figure's y-axis scaling makes small gains look larger than they are; consider adding error bars or a table-only presentation.
  5. [Appendix E, Table 13/14] The silver-label distributions are interesting, but the paper does not report any validation of Gemini-3.1-Pro's silver labels against human labels (except for the conditional gold-set prompt). A small human-evaluation sample of the silver set would help users calibrate trust in the auxiliary resource.

Circularity Check

0 steps flagged

No significant circularity: human gold labels ground the benchmark; silver labels are auxiliary and conditioned on gold.

full rationale

The paper's central contribution is an empirical benchmark, not a formal derivation. The 5K gold labels come from three trained native Arabic-speaking annotators (§3.4), and all primary results compare model predictions to these human labels. The only model-generated labels are the auxiliary silver set, which the authors explicitly separate from the core contribution: 'All primary results in this paper are reported on the human-verified 5K set.' Moreover, for the 5K gold memes, the metadata prompt 'conditions generation on the human binary and fine-grained labels rather than relabelling the meme' (Appendix E), so no gold label is produced by the model being benchmarked. The self-citations (e.g., ArMeme data-collection pipeline, MemeIntel) are used for context and collection methodology rather than as load-bearing evidence for the benchmark's validity; there is no uniqueness theorem or ansatz smuggled in via citation. The Gemma-3-12B pre-selection of which 5K memes to annotate (§3.3) is a legitimate sampling/external-validity concern, but it does not make the human annotations or model scores tautological, so it falls outside circularity.

Axiom & Free-Parameter Ledger

2 free parameters · 4 axioms · 0 invented entities

The central contribution is a measurement resource, not a derivation; no physical or formal entities are invented. The load-bearing choices are the definition of hate, the text-only inclusion filter, and the model-based pre-selection of the gold set.

free parameters (2)
  • Near-duplicate distance threshold = 3.6
    Image embeddings with Euclidean distance ≤3.6 were deduplicated as near-duplicates (§3.2); the threshold is hand-chosen and shapes which memes remain in the dataset.
  • Fine-grained sigmoid decision threshold = tuned over {0.05,...,0.90} on dev
    For multi-label hate-type prediction, the sigmoid threshold is fit to the development set (§C.2), so reported macro-F1 numbers depend on a tuned decision rule.
axioms (4)
  • domain assumption Hatefulness is defined as attack on protected characteristics (ethnicity, religion, gender, etc.); attacks on terrorist groups and generic offensiveness are not hateful.
    Established in §B.1 and applied by all annotators; these boundary rules are value-laden and define the ground truth.
  • domain assumption Memes without detectable embedded text are excluded; OCR text represents the textual modality.
    §3.2 discards such memes, so the benchmark only covers text-bearing memes and cannot generalize to purely visual or text-free memes.
  • ad hoc to paper Gemma-3-12B pre-selection produces a balanced, representative annotation pool.
    §3.3 uses a weak model to select 5K from 71K memes; no validation shows that this preserves the distribution of hatefulness or cultural diversity.
  • domain assumption The taxonomy of 7 hate types + 3 non-hate types is a complete and usable scheme for Arabic hate memes.
    The scheme is paper-defined (Table 2, §B.1); completeness is asserted, 'Other' absorbs residuals, and inter-annotator κ=0.75/0.67 suggests but does not prove full coverage.

pith-pipeline@v1.3.0-daily-deepseek · 32137 in / 11880 out tokens · 111682 ms · 2026-08-01T08:07:54.913495+00:00 · methodology

0 comments
read the original abstract

Hateful memes are a growing form of multimodal online harm, where hostile intent is often conveyed through the joint interpretation of images, text, cultural references, and implicit targets. While hateful meme detection has advanced in high-resource languages, Arabic remains underexplored, with existing meme resources focusing mainly on propaganda or coarse harmful-content labels. We introduce AHA-Memes (Arabic HAteful Memes), which is, to our knowledge, the first large-scale Arabic hateful meme benchmark with fine-grained, multi-label annotations. The dataset includes 5K manually annotated memes using a taxonomy that captures hate types, i.e., attack strategies. We further provide ~66K silver-labeled memes to support future studies. We benchmark text-only, image-only, and late-fusion multimodal models, as well as few-shot in-context learning (ICL) and open- and closed-weight Vision-Language Models (VLMs) under zero-shot and fine-tuning settings. Our results establish strong baselines and highlight key challenges in culturally grounded Arabic hateful meme detection. We release the dataset, annotation guidelines, and evaluation scripts to support future research. WARNING: This paper contains examples that may be disturbing to readers.

Figures

Figures reproduced from arXiv: 2607.27393 by Abul Hasnat, Ali Ezzat Shahroor, Firoj Alam, Md. Rafiul Biswas, Mohamed Bayan Kmainasi, Wajdi Zaghouani.

Figure 1
Figure 1. Figure 1: Examples from the AHA-MEMES dataset across hateful and non-hateful categories. shared background knowledge, rather than by ei￾ther modality in isolation (Sharma et al., 2022). This dependence on cultural context also compli￾cates human judgment, reduces label consistency, and makes automated detection more challenging. In a cross-country study, Bui et al. (2025) found that annotators from five countries ag… view at source ↗
Figure 2
Figure 2. Figure 2: Overview of the AHA-MEMES construction pipeline. cused primarily on propaganda rather than hate. The ArMeme dataset (Alam et al., 2024b) and ArAIEval shared tasks (Hasanain et al., 2024b) introduced Arabic meme propaganda-technique de￾tection, followed by work on coarse hate labels, explanation-enhanced modelling, and explainable Arabic-English propaganda detection (Alam et al., 2024a; Kmainasi et al., 202… view at source ↗
Figure 3
Figure 3. Figure 3: Few-shot learning for Qwen3-VL-8B. Binary [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Example memes annotated as Hateful / Dehumanization [PITH_FULL_IMAGE:figures/full_fig_p015_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Example memes annotated as Hateful / Inferiority. C.2 Training and Compute We fine-tune the text, image, and fusion mod￾els with the HuggingFace Transformers Trainer. We use AdamW with weight decay 0.01, a linear learning-rate schedule, 6% warmup, and 128-token text inputs. Hyperparameters are selected from the search space as reported in [PITH_FULL_IMAGE:figures/full_fig_p015_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Example memes annotated as Hateful / Incitement. Split Mean Median Max Train 85.5 76 641 Dev 87.8 80 619 Test 87.5 77 644 [PITH_FULL_IMAGE:figures/full_fig_p016_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Example memes annotated as Hateful / Mocking. Model Prompt Binary Fine-grained Gemini-2.5-pro DEFAULT 0.711 0.340 COT-EN 0.709 0.358 COT-AR 0.726 0.333 GPT-5 DEFAULT 0.628 0.301 COT-EN 0.642 0.313 COT-AR 0.647 0.315 Gemini-3.5-flash DEFAULT 0.499 0.271 COT-EN 0.499 0.272 COT-AR 0.493 0.265 [PITH_FULL_IMAGE:figures/full_fig_p017_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Example memes annotated as Hateful / Contempt. Binary (macro-F1) Fine-grained (macro-F1) Strategy K0 K1 K3 K5 K1 K3 K5 zero-shot 0.643 — — — — — — random — 0.547 0.637 0.648 0.191 0.197 0.196 text — 0.579 0.659 0.688 0.225 0.215 0.215 image — 0.577 0.648 0.690 0.219 0.210 0.213 RRF — 0.574 0.669 0.708 0.229 0.230 0.241 [PITH_FULL_IMAGE:figures/full_fig_p018_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Example memes annotated as Hateful / Slurs. for manual annotation and use the remaining memes for silver-label annotation. Prompts are listed in Section § E.1. Silver-label distributions. In Tables 13 and 14, we summarize the Gemini-3.1-Pro silver labels for the ∼66K auxiliary memes. The silver set has a low hateful rate of 0.87%. It is dominated by Egyp￾tian Arabic and MSA, everyday and entertainment topi… view at source ↗
Figure 10
Figure 10. Figure 10: Example annotated as Hateful / Exclusion. content merely for being crude, political, or about a sensitive topic. - Be objective and consistent, and apply the definitions above exactly. - Output ONLY the requested JSON -- no explanations, no extra text. # ===== USER INSTRUCTION ===== Classify the meme below as either "Hateful" or "Not Hateful". You are given: - The meme IMAGE (attached as input). - The OCR… view at source ↗
Figure 11
Figure 11. Figure 11: Example memes annotated as Not-Hateful / Humor. definitions exactly. - Output ONLY the requested JSON -- no explanations, no extra text. # ===== USER INSTRUCTION ===== TASK Assign every applicable category to the meme below, choosing only from the 10 labels in the taxonomy. You are given: - The meme IMAGE (attached as input). - The OCR-extracted Arabic text overlaid on the meme: """ {text} """ Taxonomy (l… view at source ↗
Figure 12
Figure 12. Figure 12: Example memes annotated as Not-Hateful / Sarcasm. specializing in Arabic and multilingual memes. Your task is to classify the meme AND generate additional metadata. OCR_text: {TEXT} Important: * Analyze both visual and textual content. * Read all text directly from the image. * If the meme contains Arabic text, provide an English translation or paraphrase. * Determine whether propaganda is present. * If p… view at source ↗
Figure 13
Figure 13. Figure 13: Best F1 (max over K ∈ {1, 3, 5}) per re￾trieval strategy for Qwen3-VL-2B vs. 8B. The 8B wins at every strategy; the gain from random to RRF is larger for the 8B, on both Binary (top) and Subtype (bottom). * high * very_high --- PROPAGANDA TECHNIQUES 1. name_calling_labeling 2. reductio_ad_hitlerum 3. casting_doubt 4. appeal_to_hypocrisy 5. smears 6. flag_waving 7. appeal_to_authority 8. bandwagon 9. appea… view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

23 extracted references · 3 linked inside Pith

  1. [1]

    name_calling_labeling

  2. [2]

    reductio_ad_hitlerum

  3. [3]

    * Determine whether meaning depends on text, image, or both

    Analyze multimodal meaning: * Explain how text and image interact. * Determine whether meaning depends on text, image, or both

  4. [4]

    Assign label and subtype following the classification rules above

  5. [5]

    * If no techniques are present, propaganda must be false

    Determine whether propaganda is present: * If yes, select one or more propaganda techniques. * If no techniques are present, propaganda must be false

  6. [6]

    fish season

    Large language models for data annotation and synthesis: A survey. InProceedings of the 2024 Con- ference on Empirical Methods in Natural Language Processing, pages 930–957, Miami, Florida, USA. Association for Computational Linguistics. Jianwei Wang, Tianyin Wang, and Ziqian Zeng. 2024. On the use of silver standard data for zero-shot clas- sification ta...

    Pith/arXiv arXiv 2024

  7. [7]

    label":

    Generate concise rationales: * English rationale: maximum 100 words. * Arabic rationale: maximum 100 words. * Reference visual and textual evidence when relevant. * The rationale must justify both the label and subtype assignments. Return valid JSON only. { "label": "Hateful", "subtype": "Contempt", "topic": [], "mentioned_categories": [], "dialect": "unk...

  8. [10]

    appeal_to_fear_prejudice

  9. [14]

    causal_oversimplification

  10. [15]

    black_and_white_fallacy

  11. [17]

    Analyze the visual content: * People, groups, public figures * Objects, symbols, flags, logos * Expressions, gestures, actions * Composition and visual style

  12. [18]

    thought_terminating_cliche

  13. [19]

    appeal_to_time_pressure

  14. [20]

    * Translate or paraphrase it into English

    Read and analyze all visible text: * Extract the exact text. * Translate or paraphrase it into English

  15. [21]

    obfuscation_vagueness_confusion

  16. [22]

    exaggeration_minimisation

  17. [23]

    Identify: * Topic(s) * Mentioned protected-group categories * Dialect * Cultural references * Current-event dependence * Intent * Stance * Emotional tone

  18. [24]

    other --- Instructions:

  19. [158]

    Wissam Antoun, Fady Baly, and Hazem Hajj

    IEEE. Wissam Antoun, Fady Baly, and Hazem Hajj. 2020. Arabert: Transformer-based model for arabic lan- guage understanding. InLREC 2020 Workshop Lan- guage Resources and Evaluation Conference 11–16 May 2020, page 9. Hangbo Bao, Li Dong, Songhao Piao, and Furu Wei

    2020

  20. [390]

    Firoj Alam, Abul Hasnat, Fatema Ahmad, Md

    Springer. Firoj Alam, Abul Hasnat, Fatema Ahmad, Md. Arid Hasan, and Maram Hasanain. 2024b. ArMeme: Pro- pagandistic content in Arabic memes. InProceed- ings of the 2024 Conference on Empirical Methods in 9 Natural Language Processing, pages 21071–21090, Miami, Florida, USA. Association for Computational Linguistics. Firoj Alam, Hamdy Mubarak, Wajdi Zagho...

    2024

  21. [2022]

    InInternational Conference on Learning Repre- sentations

    BEiT: BERT pre-training of image transform- ers. InInternational Conference on Learning Repre- sentations. Tom B. Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared Kaplan et al. 2020. Language mod- els are few-shot learners. InAdvances in Neural Information Processing Systems 33 (NeurIPS), pages 1877–1901. Minh Duc Bui, Katharina von der Wense, and...

    Pith/arXiv arXiv 2020

  22. [2024]

    InProceedings of the 18th Conference of the European Chapter of the Associa- tion for Computational Linguistics (Volume 1: Long Papers), pages 487–520, St

    LAraBench: Benchmarking Arabic AI with large language models. InProceedings of the 18th Conference of the European Chapter of the Associa- tion for Computational Linguistics (Volume 1: Long Papers), pages 487–520, St. Julian’s, Malta. Associa- tion for Computational Linguistics. Marah Abdin, Jyoti Aneja, Hany Awadalla, Ahmed Awadallah, Ammar Ahmad Awan et...

    Pith/arXiv arXiv 2024

  23. [2026]

    10 Maram Hasanain, Fatema Ahmad, and Firoj Alam

    Fanar 2.0: Arabic generative AI stack.ArXiv preprint, arXiv:2603.16397. 10 Maram Hasanain, Fatema Ahmad, and Firoj Alam. 2024a. Can GPT-4 identify propaganda? annota- tion and detection of propaganda spans in news arti- cles. InProceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLI...

    arXiv 2024