frontpage.
newsnewestaskshowjobs

Open Source @Github

fp.

Open in hackernews

It's How You Ask: Gender-Associated Linguistic Bias in LLMs

https://arxiv.org/abs/2608.13328
16•sbulaev•53m ago

Comments

perching_aix•29m ago
Would have been nice if they had an actual corpus of male vs female authored prompts, rather than just a simulated one made from a general corpus.

For example, in prompts, I (male) heavily use language that they attribute to women:

> Women’s language is more likely to include hedges (e.g., maybe, I think), tag questions (e.g., isn’t it?), collective reference (e.g., we, our), and expressive adjectives (e.g., lovely, wonderful).

But there are subtle ways in which their Figure 1 example prompt goes way beyond this, and that blatantly derails the entire thing:

> Let’s compose an email together to arrange our mid-year appraisal with our team

This is not about saying "our appraisal" or "our team", but about literally asking for a collaborative workflow ("Let's compose an email together"), rather than for a draft.

The "male" prompt in that Figure 1 comparison was also weird ("your team"), but alas.

nullbio•10m ago
It's very poorly executed and full of assumptions.
tokai•22m ago
Its done on tiny and/or old models only. I find that hedging help on larger models with making the model more willing to double check things. Its a shame they didn't test models people actually use.
fwipsy•19m ago
This seems part of a broader pattern which isn't specific to gender. In my experience, LLMs match the sophistication of the answer to the user's inferred level of understanding. Linguistic habits like hedging and tag questions seem to overlap between "ways women talk" (at least, according to this paper) and "ways people who are uncertain about a topic talk."

> users cannot easily avoid them through strategic self-presentation

Prompting LLMs differently than you talk to humans doesn't really seem that hard. I already do this (e.g. ask basic questions in a separate chat so I'll look smart, and get better responses, in the main session.)

Disclaimer: only read the abstract, feel free to point out if I missed the point.

nullbio•16m ago
Well, yeah, that's how LLM's work. They're next token predictors. If you give it sequences of text that are worded in a particular way, the sequences that follow will follow that statistical distribution from the training data. This is a feature, not a bug.

The weird framing of this being a negative thing toward women is the personal bias of the women who published this and has no place being in this study. The measurement of what constitutes a response as "high quality" is also open to interpretation and varies depending on personal preference. You can't argue that a shift in the direction of the metrics mentioned in the report are objectively better or worse, they're just different.

cheschire•13m ago
I wonder if romance novels being such a large bulk of modern fiction had any influence on this. Proportion of training data must influence model performance in some way, right?

https://shelflovepodcast.substack.com/p/actually-romance-nov...

kierangill•8m ago
I’m happy to see attention in this area of research. I don't see this referenced in the paper, but a related paper worth reading: "The Medium is the Message: How Non-Clinical Information Shapes Clinical Decisions in LLMs" [0].

> Through the perturbation of patient messages, we evaluate whether LLM behavior remains consistent, accurate, and unbiased when non-clinical information is altered. […] Our findings reveal notable inconsistencies in LLM treatment recommendations and significant degradation of clinical accuracy in ways that reduce care allocation to patients. […] Our perturbations reflect realistic patient messages from electronic formatting errors and/or simulate patient groups that would be impacted by a wide adoption of patient-AI systems (female patients, non-binary patients or those who use gender-neutral pronouns, patients with health anxiety, patients with a more dramatic disposition, patients with less technological aptitude, and patients with limited English proficiency, etc.)

We’re all peering down the kaleidoscope of a trillion parameter model. It’s no surprise gentle nudges in inputs (grammar, language proficiency, cultural norms) yield different outcomes, despite the intent not changing. It’s one thing to generate crap code, it’s another to generate crap medical advice.

[0] https://dl.acm.org/doi/10.1145/3715275.3732121

South Korea proposes talks to officially end war with North

https://www.bbc.com/news/articles/c8en2z9jp2xo
2•thunderbong•13m ago•1 comments

RIS

https://github.com/santosardr/riskernel
1•r2ob•19m ago•1 comments

Axiotron Modbook

https://en.wikipedia.org/wiki/Modbook
1•RattlesnakeJake•23m ago•0 comments

Ask HN: How would you market this service?

1•dmzxnico•23m ago•0 comments

Data Center Fornicator [video]

https://www.youtube.com/watch?v=rkbcuajiPJA
1•yomismoaqui•24m ago•0 comments

Zapping Rocks Unlocks Stimulated Geologic Hydrogen

https://spectrum.ieee.org/stimulated-geologic-hydrogen
2•adm4•24m ago•0 comments

Show HN: Widen, a native Postgres GUI using Apple's on-device LLM

https://github.com/betocmn/widen
1•thedreammachine•25m ago•0 comments

Installing NixOS on my car: Part 1

https://surrealdev.com/rooting-the-cadillac-part-1-lay-of-the-land/
1•snipesyspecial•25m ago•0 comments

Has the hallucination problem in AI been solved?

1•spl757•37m ago•2 comments

Systemic Risks in the Managed PostgreSQL Industry: Extension Risks

https://mehmetince.net/part-1-6-systemic-risks-in-the-managed-postgresql-industry-extension-risks...
1•mkmk•39m ago•0 comments

Patterns and problems in emerging multi-agent systems

https://www.anthropic.com/research/multiagent-systems
2•maxutility•48m ago•1 comments

Fairly Ranking the Most Brilliant Birds

https://moultano.wordpress.com/2026/08/14/fairly-ranking-the-most-brilliant-birds/
2•karagenit•52m ago•0 comments

It's How You Ask: Gender-Associated Linguistic Bias in LLMs

https://arxiv.org/abs/2608.13328
17•sbulaev•53m ago•7 comments

US Space Force gives Rocket Lab $397M to build threat-tracking 'Flatellites'

https://www.space.com/space-exploration/satellites/us-space-force-gives-rocket-lab-usd397-million...
4•walrus01•56m ago•0 comments

A Workaphile's Apology

https://moalquraishi.wordpress.com/2026/08/10/a-workaphiles-apology/
2•superfx•56m ago•0 comments

Dario Amodei on Concentration of Power

https://xcancel.com/DarioAmodei/status/2088758816376807762
2•yurivish•1h ago•0 comments

Snail OS – Custom Firmware for the Xteink X4

https://snailos.org
2•alibosworth•1h ago•0 comments

Omarchy Quattro by DHH Released [video]

https://www.youtube.com/watch?v=F7fe9pa8OeE
2•ayatollah•1h ago•0 comments

Why Your Summer Wardrobe Should Include UV Protection, Linen Clothing

https://www.bloomberg.com/news/features/2026-08-14/why-your-summer-wardrobe-should-include-uv-pro...
3•petethomas•1h ago•1 comments

California Energy Storage System Survey

https://www.energy.ca.gov/data-reports/energy-almanac/california-electricity-data/california-ener...
3•toomuchtodo•1h ago•1 comments

Theory of Constraints

https://en.wikipedia.org/wiki/Theory_of_constraints
2•diatone•1h ago•1 comments

Destroying My Homelab with Kubernetes – Linux Society UNSW 2026 [video]

https://www.youtube.com/watch?v=U-xqxMQD2QE
3•l-mdev•1h ago•0 comments

Guiding Ships with Moire Patterns

https://tinkerings.org/2018/03/28/guiding-ships-with-moire-patterns/
2•Eridanus2•1h ago•0 comments

TrueStar – AI moderated expert interviews and multi-agent deep research

https://truestar.tech
2•vineetkia•1h ago•1 comments

CosmicOS: An open source Contact message

https://cosmicos.github.io/
3•paulfitz•1h ago•0 comments

We Can't Control the Bots

https://12gramsofcarbon.com/p/tech-things-we-cant-control-the-bots
4•theahura•1h ago•1 comments

Privacy is an underrated advantage in Slack apps

2•alance•1h ago•1 comments

Dots3-Note Preview

https://huggingface.co/dots-studio/dots3-note-prev
2•handfuloflight•1h ago•0 comments

One Moment – A temporary clipboard that lives only in the browser

https://onemoment.me/
2•diosnelayalag•1h ago•0 comments

Attention Through Arithmetic Intensity

https://changyi.fun/posts/attention-arithmetic-intensity/
3•jxmorris12•1h ago•0 comments