Hacker News: betula_ai

New comment by betula_ai in "Show HN: Use Claude's projects feature with any LLM"

betula_ai — Fri, 21 Feb 2025 20:20:05 +0000

Very timely. One thing I struggle with these days, is even my starred chats are way too many. My titles aren't sufficient - would be helpful to add search. You mention "easy to find your conversation" in your site - perhaps you mean search (I haven't registered yet).

New comment by betula_ai in "Show HN: Benchmarking VLMs vs. Traditional OCR"

betula_ai — Fri, 21 Feb 2025 19:20:12 +0000

Thank you for sharing this. Some of the other public models that we can host ourselves may perform in practice better than the models listed - e.g. Qwen 2.5 VL https://github.com/QwenLM/Qwen2.5-VL?tab=readme-ov-file

New comment by betula_ai in "I think Yann Lecun was right about LLMs (but perhaps only by accident)"

betula_ai — Fri, 21 Feb 2025 18:50:35 +0000

Thank you for this informative and thoughtful post. An interesting twist to the increasing error accumulation as autoregressive models generate more output, is the recent success of language diffusion models for predicting multiple tokens simultaneously. They have a remasking strategy at every step of the reviser process, that masks low confidence tokens. Regardless your observations perhaps still apply. https://arxiv.org/pdf/2502.09992