Post

📘 Literary translation isn't just about accuracy, but also creatively conveying meaning across languages. But LLMs prompted for MT are very literal. Prompting & steering to the rescue! Can we personalize LLM’s MT when few examples are available, without further tuning? 🔍 2/

📈 But can models recognize and replicate individual translator styles?: ✓ Classifiers can find styles with high acc. (humans kinda don’t) ✓ Multi-shot prompting boosts style a lot ✓ We can detect strong style traces in activations (esp. mid layers) 3/

💡 We compare prompting (zero and multi-shot + explanations) and inference-time interventions (ActAdd, REFT and SAEs). Following SpARE (@yuzhaouoe.bsky.social @alessiodevoto.bsky.social), we propose ✨ contrastive SAE steering ✨ with mutual info to personalize literary MT by tuning latent features 4/

I love and hate this at the same, bit of double think. On one hand this is a very much needed improvement to translations but on the other hand human translators already get paid so low and this is just another excuse for companies to pay them less.