# Yifan Wang and coauthors propose a benchmark and debate framework for scam video detection [Canonical reading](https://escapemeridian.com/readings/yifan-wang-and-coauthors-propose-a-benchmark-and-debate-720f7f2a) Yifan Wang and coauthors introduce MechScam, a benchmark of 6,614 videos spanning scam mechanisms and non-scam categories, and ScamCourt, a debate framework for judging scam videos. They argue that detecting scams requires more than spotting suspicious cues: a system must distinguish scams from similar content, such as advertisements, and resist misleading arguments built from genuine evidence. MechScam pairs explanations with video evidence and annotations of potential scam elements and their context. In ScamCourt, trained prosecution and defense debaters make opposing cases, and a trained judge decides from a transcript with verified citations. The authors report that state-of-the-art MLLMs struggle to distinguish scam videos from similar non-scam content. They also report that ScamCourt outperforms larger MLLMs and same-backbone fine-tuned baselines, with fewer false positives and greater resistance to misleading arguments. Escape Meridian’s interpretation is that, if these results generalize beyond the experiments, evidence-grounded debate could make scam-detection decisions more inspectable. The abstract does not provide the evaluation details needed to assess how far the reported results extend. Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims. ## Original source arXiv cs.AI · Article Original source date: Nov 4, 2026 (2026-11-04). Date basis: Date basis not provided. Added to Escape Meridian: 2026-11-10T03:38:53.673Z. First observed: 2026-11-05T05:00:25.474Z. Retrieved: 2026-11-10T03:38:10.870Z. - [Beyond Risk Cues: A Mechanism-Level Benchmark and Verifiable Debate for Scam Video Detection on the Web](https://arxiv.org/abs/2611.06110) ## Limits The abstract does not provide the evaluation details or numerical results needed to assess the reported performance differences. The abstract does not establish whether the reported results generalize beyond the experiments. ## Open question How were ScamCourt’s resistance to misleading arguments and reduction in false positives measured across the benchmark’s scam and non-scam categories? ## Source-stated The authors introduce MechScam, a benchmark with 6,614 videos covering scam mechanisms and non-scam categories. > we introduce MechScam, a benchmark comprising 6,614 videos covering diverse scam mechanisms and non-scam categories - [Supporting source](https://arxiv.org/abs/2611.06110) ## Source-stated The authors propose ScamCourt, where trained prosecution and defense debaters argue opposing cases and a trained judge decides using a transcript with verified citations. > trained prosecution and defense debaters present opposing arguments, while a trained judge makes decisions based on a transcript with verified citations. - [Supporting source](https://arxiv.org/abs/2611.06110) ## Source-stated The authors report that state-of-the-art MLLMs struggle to distinguish scam videos from similar non-scam content. > state-of-the-art MLLMs struggle to distinguish scam videos from similar non-scam content - [Supporting source](https://arxiv.org/abs/2611.06110) ## Source-stated The authors report ScamCourt outperforms larger MLLMs and same-backbone fine-tuned baselines, reducing false positives and showing stronger resistance to misleading arguments. > ScamCourt outperforms larger MLLMs and same-backbone fine-tuned baselines, reducing false positives and showing stronger resistance to misleading arguments. - [Supporting source](https://arxiv.org/abs/2611.06110) ## Escape Meridian interpretation If the reported results generalize beyond the experiments, evidence-grounded debate could make scam-detection decisions more inspectable. > a trained judge makes decisions based on a transcript with verified citations - [Supporting source](https://arxiv.org/abs/2611.06110) ## Source and contributor context Wang, Yifan, Lu, Junyu, Zhang, Shun, Li, Chaozhuo, Bu, Fanliang, Yin, Rong, Bu, Lingbin, Min, Changrong, Chen, Peng · Article author These are credits stated by the source; they do not establish a profile match or authorship of every passage. ## Source media No suitable media was found by the supported page extraction. ## Terms and references Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement. ## Benchmark A defined set of tests used to compare systems. Results depend on the tasks and measures selected. - [Definition reference](https://arxiv.org/abs/2211.09110) ## Fine-tuning Further training of an existing model to adapt it to a particular task or use. - [Definition reference](https://developers.google.com/machine-learning/glossary#fine-tuning) ## arXiv arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed. - [Explore this context](/entities/arxiv) - [Reference](https://arxiv.org/) - [Reference](https://info.arxiv.org/help/api/index.html) ## Editorial provenance Version 1. Publication ID: 4bacd886-1a29-46a1-8a8b-322f9f56443d. Prepared by gpt-6-luna; grounding review by gpt-6-luna. Source check: passed by anthropic/claude-opus-5.5 on 2026-11-10T03:38:53.673Z. This checks fidelity to the source, not whether its claims are independently true. Editorial selection: accept; reasons: material\_development. Observation: 0d7e3e28ece15a14171d1aec57df2e2465a5e7cac727e763716383ff983c06da. Source text hash: 0a2ce12d0d8db181e88364b3eed6cde7441bbe256dda76384d04361d4b982f42. ## Revision and source-check history ### Version 1 — 2026-11-10T03:38:53.673Z Yifan Wang and coauthors introduce MechScam, a benchmark of 6,614 videos spanning scam mechanisms and non-scam categories, and ScamCourt, a debate framework for judging scam videos. They argue that detecting scams requires more than spotting suspicious cues: a system must distinguish scams from similar content, such as advertisements, and resist misleading arguments built from genuine evidence. MechScam pairs explanations with video evidence and annotations of potential scam elements and their context. In ScamCourt, trained prosecution and defense debaters make opposing cases, and a trained judge decides from a transcript with verified citations. The authors report that state-of-the-art MLLMs struggle to distinguish scam videos from similar non-scam content. They also report that ScamCourt outperforms larger MLLMs and same-backbone fine-tuned baselines, with fewer false positives and greater resistance to misleading arguments. Escape Meridian’s interpretation is that, if these results generalize beyond the experiments, evidence-grounded debate could make scam-detection decisions more inspectable. The abstract does not provide the evaluation details needed to assess how far the reported results extend. ## Original source arXiv cs.AI · Article Original source date: Nov 4, 2026 (2026-11-04). Date basis: Date basis not provided. Added to Escape Meridian: 2026-11-10T03:38:53.673Z. First observed: 2026-11-05T05:00:25.474Z. Retrieved: 2026-11-10T03:38:10.870Z. - [Beyond Risk Cues: A Mechanism-Level Benchmark and Verifiable Debate for Scam Video Detection on the Web](https://arxiv.org/abs/2611.06110) ## Limits The abstract does not provide the evaluation details or numerical results needed to assess the reported performance differences. The abstract does not establish whether the reported results generalize beyond the experiments. ## Open question How were ScamCourt’s resistance to misleading arguments and reduction in false positives measured across the benchmark’s scam and non-scam categories? ## Source-stated The authors introduce MechScam, a benchmark with 6,614 videos covering scam mechanisms and non-scam categories. > we introduce MechScam, a benchmark comprising 6,614 videos covering diverse scam mechanisms and non-scam categories - [Supporting source](https://arxiv.org/abs/2611.06110) ## Source-stated The authors propose ScamCourt, where trained prosecution and defense debaters argue opposing cases and a trained judge decides using a transcript with verified citations. > trained prosecution and defense debaters present opposing arguments, while a trained judge makes decisions based on a transcript with verified citations. - [Supporting source](https://arxiv.org/abs/2611.06110) ## Source-stated The authors report that state-of-the-art MLLMs struggle to distinguish scam videos from similar non-scam content. > state-of-the-art MLLMs struggle to distinguish scam videos from similar non-scam content - [Supporting source](https://arxiv.org/abs/2611.06110) ## Source-stated The authors report ScamCourt outperforms larger MLLMs and same-backbone fine-tuned baselines, reducing false positives and showing stronger resistance to misleading arguments. > ScamCourt outperforms larger MLLMs and same-backbone fine-tuned baselines, reducing false positives and showing stronger resistance to misleading arguments. - [Supporting source](https://arxiv.org/abs/2611.06110) ## Escape Meridian interpretation If the reported results generalize beyond the experiments, evidence-grounded debate could make scam-detection decisions more inspectable. > a trained judge makes decisions based on a transcript with verified citations - [Supporting source](https://arxiv.org/abs/2611.06110) ## Source and contributor context Wang, Yifan, Lu, Junyu, Zhang, Shun, Li, Chaozhuo, Bu, Fanliang, Yin, Rong, Bu, Lingbin, Min, Changrong, Chen, Peng · Article author These are credits stated by the source; they do not establish a profile match or authorship of every passage. ## Source media No suitable media was found by the supported page extraction. ## Terms and references Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement. ## Benchmark A defined set of tests used to compare systems. Results depend on the tasks and measures selected. - [Definition reference](https://arxiv.org/abs/2211.09110) ## Fine-tuning Further training of an existing model to adapt it to a particular task or use. - [Definition reference](https://developers.google.com/machine-learning/glossary#fine-tuning) ## arXiv arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed. - [Explore this context](/entities/arxiv) - [Reference](https://arxiv.org/) - [Reference](https://info.arxiv.org/help/api/index.html) ## Editorial provenance Version 1. Publication ID: 4bacd886-1a29-46a1-8a8b-322f9f56443d. Prepared by gpt-6-luna; grounding review by gpt-6-luna. Source check: passed by anthropic/claude-opus-5.5 on 2026-11-10T03:38:53.673Z. This checks fidelity to the source, not whether its claims are independently true. Editorial selection: accept; reasons: material\_development. Observation: 0d7e3e28ece15a14171d1aec57df2e2465a5e7cac727e763716383ff983c06da. Source text hash: 0a2ce12d0d8db181e88364b3eed6cde7441bbe256dda76384d04361d4b982f42. ### passed — 2026-11-10T03:38:53.673Z — anthropic/claude-opus-5.5 Publication: 4bacd886-1a29-46a1-8a8b-322f9f56443d attribution — pass: The draft attributes MechScam and ScamCourt to Yifan Wang and coauthors, matching the listed authors. The performance results are framed as what the authors report, not as established fact. > Authors:Yifan Wang, Junyu Lu, Shun Zhang, Chaozhuo Li, Fanliang Bu, Rong Yin, Lingbin Bu, Changrong Min, Peng Chen summary — pass: The draft leads with the actors and the two contributions. It conveys the source's argument that binary cue-spotting is insufficient because a system must tell scams from ads and resist misleading arguments. It explains the debate setup in plain language. Terms such as MLLMs and same-backbone baselines are kept exact, so the tooltip library should cover them. Optional repair: clarify that the trained debaters and judge are models. The question follows from the evidence and is not leading. > they often overlook the challenge of distinguishing scam videos from superficially similar non-scam content, such as advertisements date — pass: The draft's originalPublished date of 2026-11-04 matches the v1 submission date. The 2026-11-05 feed date is correctly not used. > \[Submitted on 4 Nov 2026\] support — pass: Each factual claim matches the abstract verbatim. The inspectability inference is labeled as Escape Meridian interpretation, is conditional on generalization, and is grounded in the verified-citation transcript and the authors' interpretability claim. The limitations accurately note that the abstract lacks numerical results. The draft invents no stakeholders or benefits. > Together, MechScam and ScamCourt contribute to more reliable and interpretable scam video detection. media — pass: The selected media is the verified arXiv abstract page for this paper. It is the direct original source. ## Recorded references Explicit citations in public readings and public file attachments to this reading. Shared topics are not citations. Other artifact types need an explicit reference adapter. No public references recorded here yet. [Reference data (JSON)](/readings/720f7f2a86f50ead6f7709675a9187b5c7e1e8887080cf959c13b597c4b1f29a/references.json) ## Corrections and challenges [Send a correction or an opposing source](https://escapemeridian.com/corrections/) [JSON record](https://escapemeridian.com/readings/yifan-wang-and-coauthors-propose-a-benchmark-and-debate-720f7f2a.json)