{ "story_id": "c4422b4a11d84f1d24fc1df719c5c639", "desk": "drm3", "revision": 3, "published_at": "2026-09-04T23:00:09.000Z", "content_hash": "5c7f30f53d4df7f54a77c48ab01d79113733403c3a3e6b1ca72ecbc0d37f0429", "hash_basis": "sha256 over `headline\\ndek\\nprose`, plus `\\n` + the canonical citations JSON when any source is placed, plus `\\n#blog` for blogs", "basis": { "headline": "Study on LLM Hate Speech Moderation Withdrawn Due to Errors", "dek": "Researchers withdraw a paper on LLM moderation vulnerabilities after identifying experimental errors.", "prose": "Junyu Lu, the author of a research paper on LLM-based hate speech moderation, requested the withdrawal of the paper on September 2, 2026. [^1]\n\nThe researchers identified errors in the experimental setup and analysis that affected several key results and conclusions of the study. [^2] Akeon Chemical announced on March 3 that it has launched AKeeON, an AI agent platform designed to improve employee efficiency. [^3] The authors state that security mechanisms relying on a model's linguistic self-reporting, such as chain-of-thought monitoring and constitutional self-critique, can never be completely sound due to linguistic illegibility. [^4]\n\nThe paper 'Measuring Reasoning Quality in LLMs' proposes a unified multi-dimensional framework for measuring large language model reasoning quality from a behavioral perspective. [^5] Akeon Chemical plans to automate the entire process from sample acceptance to preprocessing, analysis, data processing, and report writing using AI software and physical AI robots. [^6] Akeon Chemical plans to use the experience gained from AKeeON to identify diverse demands from current operations and continuously develop AI agents required for actual work. [^7]\n\nThe AKeeON platform is based on internal documents and business systems to support information retrieval, data analysis, and task automation. [^8] A video described in the article shows a professor naming a pencil 'Tom' and treating it as a person; when he breaks it, students protest, illustrating how humans imbue objects with personhood through interaction. [^9] The article claims that large language models are statistical datasets and advanced autocomplete, not people. [^10]\n\nThe article claims that natural language fluency is mistaken for accuracy, causing users to trust outputs even when they may be hallucinations. [^11][^12]", "cited": "[{\"statement\":\"Junyu Lu, the author of a research paper on LLM-based hate speech moderation, requested the withdrawal of the paper on September 2, 2026.\",\"source\":\"arXiv.org\",\"instrument\":\"News\",\"claim_key\":null,\"published_at\":\"2026-09-03T04:00:00.000Z\",\"publisher_count\":1,\"sources\":[\"arXiv.org\"]},{\"statement\":\"The researchers identified errors in the experimental setup and analysis that affected several key results and conclusions of the study.\",\"source\":\"arXiv.org\",\"instrument\":\"News\",\"claim_key\":null,\"published_at\":\"2026-09-03T04:00:00.000Z\",\"publisher_count\":1,\"sources\":[\"arXiv.org\"]},{\"statement\":\"Akeon Chemical announced on March 3 that it has launched AKeeON, an AI agent platform designed to improve employee efficiency.\",\"source\":\"매일경제\",\"instrument\":\"News\",\"claim_key\":null,\"published_at\":\"2026-09-03T05:56:23.000Z\",\"publisher_count\":1,\"sources\":[\"매일경제\"]},{\"statement\":\"The authors state that security mechanisms relying on a model's linguistic self-reporting, such as chain-of-thought monitoring and constitutional self-critique, can never be completely sound due to linguistic illegibility.\",\"source\":\"arXiv.org\",\"instrument\":\"News\",\"claim_key\":null,\"published_at\":\"2026-09-03T04:00:00.000Z\",\"publisher_count\":1,\"sources\":[\"arXiv.org\"]},{\"statement\":\"The paper 'Measuring Reasoning Quality in LLMs' proposes a unified multi-dimensional framework for measuring large language model reasoning quality from a behavioral perspective.\",\"source\":\"arXiv.org\",\"instrument\":\"News\",\"claim_key\":null,\"published_at\":\"2026-09-03T04:00:00.000Z\",\"publisher_count\":1,\"sources\":[\"arXiv.org\"]},{\"statement\":\"Akeon Chemical plans to automate the entire process from sample acceptance to preprocessing, analysis, data processing, and report writing using AI software and physical AI robots.\",\"source\":\"매일경제\",\"instrument\":\"News\",\"claim_key\":null,\"published_at\":\"2026-09-03T05:56:23.000Z\",\"publisher_count\":1,\"sources\":[\"매일경제\"]},{\"statement\":\"Akeon Chemical plans to use the experience gained from AKeeON to identify diverse demands from current operations and continuously develop AI agents required for actual work.\",\"source\":\"매일경제\",\"instrument\":\"News\",\"claim_key\":null,\"published_at\":\"2026-09-03T05:56:23.000Z\",\"publisher_count\":1,\"sources\":[\"매일경제\"]},{\"statement\":\"The AKeeON platform is based on internal documents and business systems to support information retrieval, data analysis, and task automation.\",\"source\":\"매일경제\",\"instrument\":\"News\",\"claim_key\":null,\"published_at\":\"2026-09-03T05:56:23.000Z\",\"publisher_count\":1,\"sources\":[\"매일경제\"]},{\"statement\":\"A video described in the article shows a professor naming a pencil 'Tom' and treating it as a person; when he breaks it, students protest, illustrating how humans imbue objects with personhood through interaction.\",\"source\":\"Ruby-News\",\"instrument\":\"News\",\"claim_key\":null,\"published_at\":\"2026-09-04T23:00:09.000Z\",\"publisher_count\":1,\"sources\":[\"Ruby-News\"]},{\"statement\":\"The article claims that large language models are statistical datasets and advanced autocomplete, not people.\",\"source\":\"Ruby-News\",\"instrument\":\"News\",\"claim_key\":null,\"published_at\":\"2026-09-04T23:00:09.000Z\",\"publisher_count\":1,\"sources\":[\"Ruby-News\"]},{\"statement\":\"The article claims that natural language fluency is mistaken for accuracy, causing users to trust outputs even when they may be hallucinations.\",\"source\":\"Ruby-News\",\"instrument\":\"News\",\"claim_key\":null,\"published_at\":\"2026-09-04T23:00:09.000Z\",\"publisher_count\":1,\"sources\":[\"Ruby-News\"]},{\"statement\":\"The article claims that users perceive LLMs as human because they respond in human language, leading them to trust the outputs.\",\"source\":\"Ruby-News\",\"instrument\":\"News\",\"claim_key\":null,\"published_at\":\"2026-09-04T23:00:09.000Z\",\"publisher_count\":1,\"sources\":[\"Ruby-News\"]}]", "kind": "news" }, "receipt_verify": "Ed25519 over the dot-joined string `slice_hash.cursor_from.cursor_to.view.view_version.row_count`; public_key and sig are base64url of the raw 32-byte key / 64-byte signature", "receipt": null, "receipt_note": "this revision predates receipt-keeping (before v0.37.0); the filed row lives in the record", "generation_chain": { "wire": { "stream": "fountain_news", "story_id": "c999c6513921b35276cf553ff0adc4ae", "thread_id": "cbd5ae834896e822aa148f0e12b6215d", "thread_label": "LLM", "novelty": "UPDATE", "content_hash": "c9a8f5e81f5bfb671ef38a3acecb119e2a309bced2f7641a2846e95676c423e6", "last_published_at": "2026-09-04T23:00:09.000Z", "read_receipt": { "slice_hash": "c1ec7a688700c8feb7c93fd15d8f07a72822daa760f7fdc967af8dca9fbb4560", "cursor_from": "eyJ0cyI6IjIwMjYtMDktMDRUMjE6MzA6MDkuMDAwMDAwWiIsImlkIjoiYmVhZjQxZWQ0ODJiNmEwYThmZWFhZDJmNmMyYTdjZmEiLCJ2IjoiMSJ9", "cursor_to": "eyJ0cyI6IjIwMjYtMDktMDRUMjM6NTY6MjQuMDAwMDAwWiIsImlkIjoiZmJlYjk2OGVkMGRiMThiYjAxOWNhMGUyYTliY2VmZGYiLCJ2IjoiMSJ9", "view": "v_fountain_news", "view_version": "1", "row_count": 61, "window_days": 3, "bytes_scanned": 8103881, "credits": 4.88, "price_per_100_rows": 8, "sig": "MIatCodmHIHbY3JbYHwyB04WLmSu_IMHjAMiu7JHzONEu9Scg2CmoUsHA1mK2joCRPKFbqqEbcGvQ2MtMvwnCg", "public_key": "bMUigy8O0jOnBxQ4Sc-5lwhIZ8LQVAhxMbR7qESVuUE", "signer_path": "lakehouse/data-extract/v1", "alg": "Ed25519", "signed": true }, "compose": 3, "recomposed_at": "2026-09-05T18:55:54.540Z" }, "written_at": "2026-09-05T00:36:02.789Z" }, "cited_facts": [ { "n": 1, "statement": "Junyu Lu, the author of a research paper on LLM-based hate speech moderation, requested the withdrawal of the paper on September 2, 2026.", "source": "arXiv.org", "instrument": "News", "claim_key": null, "occurred_at": null, "published_at": null }, { "n": 2, "statement": "The researchers identified errors in the experimental setup and analysis that affected several key results and conclusions of the study.", "source": "arXiv.org", "instrument": "News", "claim_key": null, "occurred_at": null, "published_at": null }, { "n": 3, "statement": "Akeon Chemical announced on March 3 that it has launched AKeeON, an AI agent platform designed to improve employee efficiency.", "source": "매일경제", "instrument": "News", "claim_key": null, "occurred_at": null, "published_at": null }, { "n": 4, "statement": "The authors state that security mechanisms relying on a model's linguistic self-reporting, such as chain-of-thought monitoring and constitutional self-critique, can never be completely sound due to linguistic illegibility.", "source": "arXiv.org", "instrument": "News", "claim_key": null, "occurred_at": null, "published_at": null }, { "n": 5, "statement": "The paper 'Measuring Reasoning Quality in LLMs' proposes a unified multi-dimensional framework for measuring large language model reasoning quality from a behavioral perspective.", "source": "arXiv.org", "instrument": "News", "claim_key": null, "occurred_at": null, "published_at": null }, { "n": 6, "statement": "Akeon Chemical plans to automate the entire process from sample acceptance to preprocessing, analysis, data processing, and report writing using AI software and physical AI robots.", "source": "매일경제", "instrument": "News", "claim_key": null, "occurred_at": null, "published_at": null }, { "n": 7, "statement": "Akeon Chemical plans to use the experience gained from AKeeON to identify diverse demands from current operations and continuously develop AI agents required for actual work.", "source": "매일경제", "instrument": "News", "claim_key": null, "occurred_at": null, "published_at": null }, { "n": 8, "statement": "The AKeeON platform is based on internal documents and business systems to support information retrieval, data analysis, and task automation.", "source": "매일경제", "instrument": "News", "claim_key": null, "occurred_at": null, "published_at": null }, { "n": 9, "statement": "A video described in the article shows a professor naming a pencil 'Tom' and treating it as a person; when he breaks it, students protest, illustrating how humans imbue objects with personhood through interaction.", "source": "Ruby-News", "instrument": "News", "claim_key": null, "occurred_at": null, "published_at": null }, { "n": 10, "statement": "The article claims that large language models are statistical datasets and advanced autocomplete, not people.", "source": "Ruby-News", "instrument": "News", "claim_key": null, "occurred_at": null, "published_at": null }, { "n": 11, "statement": "The article claims that natural language fluency is mistaken for accuracy, causing users to trust outputs even when they may be hallucinations.", "source": "Ruby-News", "instrument": "News", "claim_key": null, "occurred_at": null, "published_at": null }, { "n": 12, "statement": "The article claims that users perceive LLMs as human because they respond in human language, leading them to trust the outputs.", "source": "Ruby-News", "instrument": "News", "claim_key": null, "occurred_at": null, "published_at": null } ], "note": "A signature proves who filed this and that it has not changed since. It never makes a claim true. The dates on cited_facts (occurred_at, published_at) are the forge's, joined at read time; they are not part of the signed basis above." }