Show HN: Evaluating Local LLMs as language translators for my app

4•3stacks•2h ago

This is my first attempt at running an eval of this nature so would love some methodology feedback.

I can't guarantee the sources weren't already in the model's inputs without getting novel translations from native speakers, but from my experience using the top models, they feel very accurate. Even encountering somewhat obscure texts from a relatively small language the translations generally beat Google Translate for proper idiomatic meaning.

Comments

robby1110•2h ago

Interesting study! First time I've really looked at this sort of study so probably not the best to comment on methodology but the overall premise of using LLMs for language conversion to increase an apps TAM is appealing

3stacks•2h ago

Thank you! Running phrase translations through an LLM really helps with processing idiomatic phrasing, and also enables quite good output practice. Spoken output practice would be a game changer.

There's definitely a lot of potential in this space especially in generating learner content in languages where there's sparse content online.

Baseball Coaches or Parents

Entertainmentindustry.ai – Engineering hunt and M&A gateway

You can't train in what the model knows

Why isn't xAI as impactful compared to others?

Tree Transformers

LLMs Are Complicated Now

A New Book on Plagiarism

Animaps: An open source travel map video maker

Why has the pointe shoe been so resistant to change?

How to Feed a Dictator

'Digital Circus' Google easter egg

Statement regarding GNU Savannah security reports

A record 242 US cities now have starter homes that cost $1M

Blackwell Build Atlas

I am dreading our LLM-written incident report future

Build a full knowledge graph locally with any video

I Solved My Mystery Fatigue with AI

Why Algo Traders Club Is Moving from Hyperliquid to Kalshi

Telo confirms 400 kW sustained charging for its tiny electric truck

Efficient C++ Programming for Modern C++ CPUs, Chapter 4/part 2

Out of Stealth (Kinda)

Kahneman: Theory Induced Blindness

Show HN: Agentcard – virtual cards for AI agents, now with DoorDash checkout

How to avoid street violence, by a professional kickboxer

Show HN: P34 – tabular profit prediction without the optimism bias

Lump Sum vs. DCA comparison: LS wins or does it?

Cursor Is Now SpaceX: Enterprise Agentic Coding's New Lock-In Risk

Backstitch: Real-Time Version Control for Godot

Swedish VPN provider OVPN sold to US stock and asset company

We built a lab to evaluate data agents – Hex