frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

fp.

Open in hackernews

LLMs Don't Hallucinate – They Drift

https://figshare.com/articles/conference_contribution/Measuring_Fidelity_Decay_A_Framework_for_Semantic_Drift_and_Collapse/30422107?file=58969378
16•knowledgeinfra•1h ago

Comments

knowledgeinfra•1h ago
This paper argues that the dominant metaphor for LLM failure, hallucinations, misdiagnoses the real problem. Language models do not primarily fail by inventing false facts, but by undergoing fidelity decay, the gradual erosion of meaning across recursive transformations. Even when outputs remain accurate and coherent, nuance, metaphor, intent, and contextual ground steadily degrade. The paper proposes a unified framework for measuring this collapse through four interrelated dynamics, lexical decay, semantic drift, ground erosion, and semantic noise, and sketches how each can be operationalized into concrete benchmarks. The central claim is that accuracy alone is an insufficient evaluation target. Without explicit fidelity metrics, AI systems risk becoming fluent yet hollow, technically correct while culturally and semantically impoverished.
petesergeant•58m ago
Please don’t post AI summaries here
chrisjj•52m ago
> Language models do not primarily fail by inventing false facts, but by undergoing fidelity decay

This premise is unsound. We don't expect LLMs to deliver with fidelity, just as we don't expect parrots to speak with their owners' accents. So infidelity is by no means a failure.

zahrevsky•1h ago
> The contribution of this work lies in its move from critique to measurement. It proposes concrete methods: recursive summarization chains, metaphor stress-tests, resonance surveys, and noise-infused retrieval experiments. These allow researchers to track how meaning erodes over time. By integrating these methods, it outlines a pathway toward fidelity-centered benchmarks that complement existing accuracy metrics.

To me, starting to solve the problem by meticulously measuring it, is a sign of a good solution.

Retr0id•55m ago
What the heck is a resonance survey
chrisjj•51m ago
An LLM fabrication.
chrisjj•1h ago
True title: Measuring Fidelity Decay: A Framework for Semantic Drift and Collapse
botacode•1h ago
Getting a 403 when I try to read. Anyone have a backup link?
Retr0id•53m ago
This is slop
sylware•50m ago
ofc not, they "bungee jump"

:p

m0llusk•30m ago
Hallucinations that have certain characteristics and boundaries are still hallucinations. This is happening because learning models are doing pattern matching, so to put it briefly anything that fits may work and end up in the output.

Being able to admit the flaws and limitations of a technology is often critical to advancing adoption. Unfortunately, producers of currently popular learning model based technologies are more interested in speculation and growth and speculative growth than genuinely robust operation. This paper is a symptom of a larger problem that is contributing to the bubble pop, downturn, or "AI winter" that we are collectively heading toward.

polotics•6m ago
This is so short and empty sorry, the author would be well placed to try to ground their work in a modicum of empiricism, the puffed-up style here makes things a bit hard to read. I do not know if this is slop it's getting harder to guess, and some actual humans have been writing like this long before LLMs. Still, what is the actual finding being presented here?

Iran Protest Death Toll Could Top 30k, According to Local Health Officials

https://time.com/7357635/more-than-30000-killed-in-iran-say-senior-officials/
1•mhb•1m ago•0 comments

Lawsuit claims Meta can see WhatsApp chats in breach of privacy

https://finance.yahoo.com/news/lawsuit-claims-meta-see-whatsapp-013745124.html
1•phyzix5761•3m ago•0 comments

The IndieWeb and Small Web

https://christiano.dev/post/indieweb_smallweb/
1•todsacerdoti•4m ago•0 comments

What is the best way to train for a marathon?

https://www.economist.com/science-and-technology/2025/12/26/what-is-the-best-way-to-train-for-a-m...
1•rienbdj•5m ago•0 comments

Secret 'discombobulator' weapon was crucial to Venezuelan raid on Maduro

https://nypost.com/2026/01/24/us-news/trump-reveals-to-the-post-secret-discombobulator-weapon-was...
2•diogenes_atx•5m ago•0 comments

Computing Sharding with Einsum

https://blog.ezyang.com/2026/01/computing-sharding-with-einsum/
1•matt_d•6m ago•0 comments

Climber Alex Honnold scales 101-floor skyscraper without safety gear

https://www.bbc.com/news/articles/c4gl0njzxjdo
1•bookofjoe•10m ago•1 comments

Show HN: Nyola – A daily Pareidolia tool (draw what you see in clouds)

https://apps.apple.com/us/app/nyola/id6755757565
1•Foilleuse•12m ago•1 comments

Agent Skills Threat Model

https://safedep.io/agent-skills-threat-model/
1•abhisek•12m ago•0 comments

I Have Spent 500 Hours Programming With AI. This Is what I learned [video]

https://www.youtube.com/watch?v=91B_v-wOaws
1•EPendragon•13m ago•0 comments

Show HN: Interactive "Zero to Hero" – Practice what you learn with live feedback

https://zero-to-hero.app/
1•jayseb•16m ago•0 comments

Show HN: JsonUI – Constrain AI agents through code structure, not prompts

1•tai-kimura•17m ago•0 comments

Ask HN: What are the most significant man-made creations to date?

2•George97•24m ago•4 comments

Free climbing in Greenland: Arctic ascent with Alex Honnold [video]

https://www.youtube.com/watch?v=ep-xRQDTiOg
1•teleforce•24m ago•0 comments

Breakmeifyoucan: Exploiting PKO and Relay Attacks in 3DES/AES NFC Technologies

https://breakmeifyoucan.com/
2•netsec_burn•24m ago•0 comments

Show HN: Local Masonry Video Player – Pinterest UI, Prompt Search, Mobile Stream

https://github.com/HoujyouChomei/local-masonry-video-player
1•choumei•25m ago•1 comments

Tinder for Issues

https://twitter.com/acolombiadev/status/2014830414410518885
1•andreag11•26m ago•0 comments

PlowNYC: Track the progress of DSNY snow removal vehicles

https://plownyc.cityofnewyork.us/plownyc/
1•exegete•26m ago•0 comments

Show HN: HomeGenGuide – Calculator for home generator installation costs

https://www.home-generator-installation.com
1•vansxxx•27m ago•0 comments

EditTools

https://edittools.org
1•zhouhua•28m ago•0 comments

Fixing Breadboards for Wide Microcontrollers – Pico and ESP32 Edition

https://www.instructables.com/Fixing-Breadboards-for-Wide-Microcontrollers-Pico-/
1•rbanffy•28m ago•0 comments

Show HN: Reminders to Stay in Touch with Friends

https://myfriends.lol
1•alabhyajindal•29m ago•0 comments

Who is using AI to code? Global diffusion and impact of generative AI

https://www.science.org/doi/10.1126/science.adz9311
4•oss_fan•33m ago•1 comments

I reverse-engineered Kindle to build on-demand AI audiobooks

https://blog.ryanbbrown.com/p/i-reverse-engineered-kindle-to-build
2•ryanbbrown•33m ago•0 comments

SoundCloud deleted 12 years of my music – so I built my own

3•miguelmichelson•33m ago•0 comments

In Praise of Pre-Hays: "Morocco" and the Public Domain

https://blog.archive.org/2026/01/20/in-praise-of-pre-hays-morocco-and-the-public-domain/
1•Kye•37m ago•0 comments

Show HN: TUI to track stock and cryptocurrencies in real-time

https://github.com/ni5arga/stock-tui
1•ni5arga•38m ago•0 comments

Post Takeover Ethics

https://gist.github.com/muratozkan/b0918e359532766abeaf9202420516e5
1•BloodRavens•39m ago•0 comments

Speed Vertigo: A New Kind of Engineering Debt

https://joshtuddenham.dev/blog/vertigo/
2•bananaboy•39m ago•0 comments

Show HN: Shorlabs – the Vercel for backend (open-source)

https://github.com/aryankashyap0/shorlabs
1•vforbackend•39m ago•0 comments