David Mimno reposted
@404media.co wrote about our new preprint on tell-tale signs of AI-generated stories! Cc @dmimno.bsky.social Paper: arxiv.org/abs/2605.26492 Article: www.404media.co/elias-thorne...
Who's Who of AI
David Mimno
Information science professor at Cornell, machine learning
Why they matter
Information science professor at Cornell, machine learning with public evidence across AI research.
- AI signals
- 6 past 30d
- Sources
- 4 distinct domains
- Discusiones
- 11 past 30d
- Latest signal
- 7d ago
He teaches information science at Cornell. http://mimno.infosci.cornell.edu
What they're sharing
Chatbots Keep Telling Stories About Lighthouse Keeper 'Elias Thorne'. We Might Know Why 404media.co
View on Bluesky →
Elias in the Lighthouse, Again? Diagnosing Low Diversity in LLM Stories arxiv.org
GitHub - NandhaKishorM/laya github.com
View on Bluesky →
convaiinnovations/laya · Hugging Face huggingface.co
aac6fef/laya-mlx · Hugging Face huggingface.co
GitHub - mizorewww/laya-mlx: Native MLX runtime for Laya typed decision models — 7–14 ms short decisions on M3 Max. No text generation, PyTorch, or cloud API. github.com
xiaoningwang (Xiaoning Wang) huggingface.co
Neutrality Bites: Gender Representation in AI-Generated Animal Stories arxiv.org
BHL OCR Leaderboard - a Hugging Face Space by finebooks huggingface.co
OCR on the Hub - a davanstrien Collection huggingface.co
arxiv.org
View on Bluesky →
arxiv.org
View on Bluesky →
Articles & links
The preprint is arxiv.org/abs/2605.26492, and @srhm.ca is working on a post!
AI Weekly's analysis
→
Read full analysis →
David Mimno reposted
This seems to be the most popular alternative to jev. Laya: Multilingual, non-autoregressive System 1 decision model. huggingface.co/convaiinnova... Laya Repo: github.com/NandhaKishor... Laya MLX: huggingface.co/aac6fef/laya... Laya MLX Repo: github.com/mizorewww/la...
David Mimno reposted
This seems to be the most popular alternative to jev. Laya: Multilingual, non-autoregressive System 1 decision model. huggingface.co/convaiinnova... Laya Repo: github.com/NandhaKishor... Laya MLX: huggingface.co/aac6fef/laya... Laya MLX Repo: github.com/mizorewww/la...
AI Weekly's analysis
→
- Convai Innovations released Laya, a 421M-parameter decision head built on ModernBERT-large, under Apache 2.0 with three checkpoints.
- p50 latency of 32.8ms on a Tesla T4 versus TypeSafe Jev's 236-276ms is the model's headline performance claim.
- Zero-shot typed-decisions accuracy is 0.362 (near random); the 0.766 figure requires fine-tuning on that benchmark's own training split.
Read full analysis →
View on Bluesky →
David Mimno reposted
This seems to be the most popular alternative to jev. Laya: Multilingual, non-autoregressive System 1 decision model. huggingface.co/convaiinnova... Laya Repo: github.com/NandhaKishor... Laya MLX: huggingface.co/aac6fef/laya... Laya MLX Repo: github.com/mizorewww/la...
AI Weekly's analysis
→
- aac6fef released laya-mlx, a native MLX FP16 port of the 0.4B-parameter convaiinnovations/laya decision encoder, running on Apple silicon without PyTorch.
- On an Apple M3 Max the port matched upstream PyTorch MPS FP32 on argmax in 63 of 63 decision distributions, with a max probability difference of 0.0054443.
- The 843 MB Apache-2.0 checkpoint handles choice, ordinal score, and boolean noul questions inside a 512-token total context.
Read full analysis →
View on Bluesky →
David Mimno reposted
This seems to be the most popular alternative to jev. Laya: Multilingual, non-autoregressive System 1 decision model. huggingface.co/convaiinnova... Laya Repo: github.com/NandhaKishor... Laya MLX: huggingface.co/aac6fef/laya... Laya MLX Repo: github.com/mizorewww/la...
AI Weekly's analysis
→
- Laya MLX reports 13.4 ms median latency for English typed decisions and 7.4 ms for the multilingual checkpoint on an M3 Max.
- The runtime skips text generation entirely, running ModernBERT-large and mmBERT-base encoders through MLX with no PyTorch or cloud dependency.
- It handles three question shapes: choice probabilities, rubric-based scores, and proposition probabilities, all local and Apple Silicon only.
Read full analysis →
View on Bluesky →
David Mimno reposted
If you've been using Talkie-1930-base or Talkie-web, and you'd like to load them simply in HF Transformers format, rather than via PyTorch checkpoint, @xiaoningwang.ca has you covered: huggingface.co/xiaoningwang
AI Weekly's analysis
→
Read full analysis →
View on Bluesky →
David Mimno reposted
If you liked Bears Will Be Boys, you might enjoy our new FAccT paper! We prompted LLMs to complete 24K stories about animal characters where gender is unstated. We found that.. bears are *still* boys. And female animal characters disappeared while "neutrality" increased. arxiv…
AI Weekly's analysis
→
- Across 23,800 AI-generated stories, feminine animal characters appeared in just 2.2% versus 40.6% masculine.
- Models avoided assigning any gender 19% of the time on average; gender-neutral 'it/its' pronouns appeared in 38.2% of stories.
- The authors argue AI neutrality can erase marginalized identities rather than protect them, challenging a common alignment assumption.
Read full analysis →
View on Bluesky →
David Mimno reposted
A 15.9M-parameter specialist OCR model outperforms many VLMs hundreds of times larger. On a 2,165-page historical eval dataset, Kraken PP-OCRv6 ranks #4 on reading CER and #1 when long-s, ligatures and case are preserved. huggingface.co/spaces/fineb...
AI Weekly's analysis
→
- rednote-hilab's dots.mocr (3B) leads the FineBooks BHL leaderboard at 97.6% reading accuracy across 2,165 expert-transcribed pages.
- Sub-2B models OvisOCR2 (0.9B, 96.9%) and PaddleOCR-VL-1.6 (1B, 96.1%) come next at under $0.50 per thousand pages.
- Ground truth covers only Antiqua typefaces in English, French, German and Latin; Fraktur, handwriting and multi-column layouts are excluded.
Read full analysis →
View on Bluesky →
David Mimno reposted
OCR for Japanese manga, Swedish handwriting or Arabic print? There’s a growing range of OCR models on @hf.co: VLMs, dedicated text recognisers and complete OCR pipelines. I’ve gathered 41 models into four collections, with short notes to help you choose: huggingface.co/collect…
AI Weekly's analysis
→
- davanstrien's OCR on the Hub groups models into four sub-collections covering documents, languages and scripts, handwriting and archives, and text recognition pipelines.
- The languages sub-collection covers Thai, Japanese, Vietnamese, Arabic, Korean and Devanagari; handwriting spans Swedish, Norwegian, German Kurrent, Tibetan, and Hebrew-script manuscripts.
- The text recognition sub-collection features Kraken and PaddleOCR pipelines for documents, manga and text in photographs.
Read full analysis →
View on Bluesky →
David Mimno reposted
✨New Work✨ forthcoming at #AIES2026: 1️⃣ "Data Annotation as Measurement" by me, @allisonkoe.bsky.social, and @kizilcec.bsky.social explores how data annotation can go wrong, and how measurement theory can help improve it. 🔗: arxiv.org/pdf/2608.07297
David Mimno reposted
This is the best thing I've read on baseball's new ABS system. Balls and strikes are made, not found. arxiv.org/pdf/2605.16237
Their own posts
Recent commentary
It's possible for Jev/Laya/Decision Models to be not that big a deal as tech and massive as a new paradigm. Here's why I'm really excited from an NLP history perspective (thread)
I’m all for serious penalties for hallucinated citations. But there’s a reason one of the rules of my lab is “Don’t try to solve Library Science”.
Cars are a surprisingly good model for public acceptance of AI. Some people love them and tinker with them. Others hate and reject them. Most find their own car useful. But everyone viscerally hates everyone else's car: traffic, parking, dangerous driving, pollution, etc.
Can anyone suggest even one single reason anyone would ever want to watch an AI generated video beyond gee-whiz novelty? Who wants this?
“Never trust an AI executive with a bunker” -Brent Hecht
I googled a quote I half remembered ("one bite from the sunny side of the peach"). No results, but I thought I'd try the AI mode. It said this was Diderot criticizing Rousseau. I asked for a source, and it pointed me to Scott's Waverley (the correct answer it seems). What happened to Diderot?
Their network
In David Mimno's orbit
Center = David Mimno. Left = members they follow (green edges). Right = members who follow them (blue edges). Top = mutual follows (orange edges, slightly larger). Drag any node to reposition; click to open that profile.
Are you David Mimno? Show it.
Add the Who’s Who of AI badge to your site or bio. It links back to this profile.
Markdown: [](https://aiweekly.co/whos-who/person/david-mimno)