#ml #article

AI Self-preferencing in Algorithmic Hiring: Empirical Evidence and Insights (PDF)

Using a large-scale controlled resume correspondence experiment, we find that LLMs consistently prefer resumes generated by themselves over those written by humans or produced by alternative models, even when content quality is controlled. The bias against human-written resumes is particularly substantial, with self-preference bias ranging from 67% to 82% across major commercial and open-source models. To assess labor market impact, we simulate realistic hiring pipelines across 24 occupations. These simulations show that candidates using the same LLM as the evaluator are 23% to 60% more likely to be shortlisted than equally qualified applicants submitting human-written resumes, with the largest disadvantages observed in business-related fields such as sales and accounting. We further demonstrate that this bias can be reduced by more than 50% through simple interventions targeting LLMs' self-recognition capabilities.

Из интересного: предпочтение собственной генерации коррелирует со способностями LLM к распознанию собственного вывода, которая, в свою очередь, коррелирует с количеством весов. Описанные авторами "simple interventions" включают в себя два способа. Первый — явное включение в промпт указание на игнорирование авторства текста. Второй — использование ансамбля LLM, включающих в себя модели с более слабыми способностями к самораспознаванию.

#ml #meme

#ml #itsec

Source

#ml #meme

#ml #meme

(наверняка троллинг, но всё равно смешно)

#prog #ml #abnormalprogramming

Llama.ttf

The font shaping engine HarfBuzz, used in applications such as Firefox and Chrome, comes with a Wasm shaper allowing arbitrary code to be used to "shape" text.

In particular, this "arbitrary" code could in principle be an entire LLM inference engine with trained parameters bundled inside, relying on treating text containing magic symbols for fake "ligatures" to initialize the LLM and use it to generate text.

It could also in principle be an entire LLM inference engine (Llama in our case, hence the name) except instead of only being in principle it's what this is.

Есть ссылка на видео с демонстрацией. Сам шрифт весит 60 мегабайт, по очевидным причинам.

(thanks @programming_sucks)

Incident Report: CVE-2026-LGTM

Необычный жанр: #itsec-сатира (и ещё немного #ml)

I think LLM is a Literal Lies Machine.

#ml #quotes

#meme про #prog-раммирование со спецификациями (и #ml-adjusted, видимо)

#prog #rust #ml

rust-lang/rust is adopting an LLM policy

TL;DR: политика формализует правила, которые уже были неофициально.

I've seen LLMs causing three main issues for our community:

* Polished technical products no longer indicate effort and understanding.
* Making code easier to write exacerbates our existing issues with review bandwidth.
* People mechanically copy-pasting to and from an LLM is a waste of our time.

As time went on, these issues grew and grew, until we had to create dedicated channels and moderation policy for how to deal with them. However, those channels work against our goals of being transparent and welcoming, because new contributors have no idea what the rules are.

The new policy formalizes those rules publicly, so that new contributors know how to join our community without getting their PRs closed for reasons they don't understand, and so that existing reviewers can easily point to the rules as an actionable reason when closing PRs that don't follow them.

#prog #zig #ml #article

Собственно, именно по этой причине в Zig запретили использование LLM вообще.

Contributor Poker and Zig's AI Ban

TL;DR: Zig foundation заинтересованы в том, чтобы контрибуторы вовлекались в сообщество и набирались экспертизы, а использование LLM очень часто на практике означает отсутствие желания разбираться и оставаться с сообществом.

#ml #suckassstory

#ml #предложка, пожалуй

#ml #meme

#ml #meme

#ml #meme

(thanks @ascii_nova_blog)

#prog #ml #rust #abnormalprogramming

slopc

A proc macro that uses an hallucination machine to write your function bodies at compile time. Throw a signature, slap #[slop] on it, and pray. If it doesn't compile, it feeds the errors back and retries until it does (or gives up).

(thanks @saddevdiary)

#ml #meme про муху