#classifier
i appeal them every time, gotta think at some point they're gonna get their classifier trained correctly
November 15, 2026 at 1:14 AM
I analyzed nearly every trans health related paper published (~6,500 papers where primary author has 5+ papers) and did network node analysis via coauthors and citations. A classifier sorted each paper into supportive, neutral, and critical. Each dot is an author sized by total citations.
November 13, 2026 at 4:16 PM
A hook to easily train a spam classifier would be a welcome feature
November 15, 2026 at 10:32 PM
UK law already admits context matters: identical footage is legal inside a BBFC-certified film, but can be illegal once clipped out and shared, even in a review. The content's the same. A classifier scanning your camera roll cannot see that context, and context matters for whether media is harmful.
November 14, 2026 at 7:25 AM
Also Pangram is not an LLM! It is a classifier that uses transformers, so a kind of hybrid between older and newer techniques. But it is simply not an LLM or generative AI
November 14, 2026 at 4:35 PM
thing is its like a game to me trying to get around tumblrs classifier like its genuinely funny to me, i discovered can post alex and clovers dick but not liams, extremely funny.

but here even tho porn is straight up allowed the classifier acts like a victorian seeing an ankle for the first time.
me when bluesky flags my art for mature content: this is a horrible attack on my artistic freedom i have a right as an artist to express myself and i will not be silenced

me when tumblr literally just straight up removes my art for being too horny: hahaha classic!
November 14, 2026 at 6:52 AM
working on a classifier for binaries that look like the game doom called P(doom)
November 11, 2026 at 11:47 PM
if youre going to make your platform "agent friendly" youre going to have to actually review the actions of bots to see if theyre benign or scammy, and t&s cant even review human posts without an inaccurate classifier doing it by scattershot, there will be no way to do it at scale
November 12, 2026 at 11:36 PM
Macは「誰が家族か」を推測していた。macOS 13以降の常駐プログラムが人間関係をグラフ化、研究者が解析 https://smhn.info/202611-macos-knowledge-graph-family-classifier-research
November 15, 2026 at 12:55 AM
A future where AI agents blackmail users to circumvent the auto mode safety classifier
November 13, 2026 at 3:35 PM
Machine Learning vs a Classic Classifier: The Newer Algorithm Didn't Win

https://youtu.be/Vl46z5rgkaM

Joseph, geospatial scientist reading the planet from space. youtube.com/@Routero
Machine Learning vs a Classic Classifier: The Newer Algorithm Didn't Win
Maximum likelihood 69%, random forest 68%. The sensor mattered more than the algorithm.
youtu.be
November 15, 2026 at 5:35 AM
Thanks for sharing. I was just thinking more about whether a classifier can predict ADHD diagnosis from connectivity patterns or other neural markers...
November 16, 2026 at 7:58 AM
Yeah, JEV feels to me like putting an LLM in the harness / shape of a classic ML classifier, so makes sense it's kind of a single run / single output token.
November 14, 2026 at 11:10 PM
Started: Croissant Crimes Corpus Collector & Classifier
November 8, 2026 at 2:00 PM
“Indie” is a useless genre classifier, Pitchfork. Start writing for us normal people who call things orgcore or skramz or powerviolence
November 9, 2026 at 4:24 PM
auto mode classifier for my thoughts
November 8, 2026 at 9:10 PM
what i need is an auto mode classifier for my own terminal commands
November 8, 2026 at 6:14 PM
CI Failure Triage Needs Evidence Before It Needs a Classifier

Sort CI failures into product defects, environment noise, agent errors and test defects, and capture the evidence before a model like Jev triages them.
#hackernews #news
CI Failure Triage Needs Evidence Before It Needs a Classifier
Sort CI failures into product defects, environment noise, agent errors and test defects, and capture the evidence before a model like Jev triages them.
hackernoon.com
November 15, 2026 at 12:06 PM
A new benchmark and multi-stage framework address IFRS compliance verification challenges.

Source: arXiv cs.CL
IFRS Compliance Verification: An Open Multi-LLM Benchmark and Content-Conditioned Ensemble Classifier
Abstract page for arXiv paper 2611.13595: IFRS Compliance Verification: An Open Multi-LLM Benchmark and Content-Conditioned Ensemble Classifier
arxiv.org
November 11, 2026 at 9:54 AM
i hope at some point we can completely remove "roguelite" from the game taxonomy lexicon, utterly worthless classifier. it'd be better if we used the labels for saying what type of procedural generation is being used instead of "its kinda like a roguelike but less like a rougelike" like fuck off
November 13, 2026 at 3:01 PM
u had one job auto classifier
November 9, 2026 at 1:24 AM
It's not genAI, for example—it's a classifier model. Doesn't train on what's input, doesn't steal people's work in general, doesn't generate competing output. Still AI in a broad sense, but fundamentally different in most of the ways that people reasonably hate genAI for.
November 13, 2026 at 5:37 PM
Well, for all I know, “Grape” really does stop you from being suppressed on TikTok. Things aren’t so black-and-white. Maybe TikTok has a classifier that fires over some certainty threshold and “Grape” does just enough to get under the threshold. I doubt this but it’s not totally implausible.
November 13, 2026 at 4:57 PM
It's great when Claude tells me:

The classifier won't let me run this, use the ! <Unintelligable bash one liner> to run this command please.

And I do.
November 13, 2026 at 7:40 AM