Maxime Rivest π§ββοΈπ¦π§
@MaximeRivest
Easy LLM context for all! β¨pip install attachments Inspired by: ggplot2, DSPy, claudette, dplyr, OpenWebUI! Follow for: API design, AI, and Data πCCππ maker
in case you are wondering this is academia now
ICMLβs Statement about subversive hidden LLM prompts We live in a weird timelineβ¦
The evals on this are quite insane! ~6B model packs quite a punch β‘
NEW: Higgs Audio V2 from @boson_ai open, unified TTS model w/ voice cloning, beats GPT 4o mini tts and ElevenLabs v2 π₯ > Trained on 10M hours (speech, music, events) > Built on top of Llama 3.2 3B > Works real-time and on edge > Beats GPT-4o-mini-tts, ElevenLabs v2 in prosodyβ¦
This seems too good to be true :O
NEW: Higgs Audio V2 from @boson_ai open, unified TTS model w/ voice cloning, beats GPT 4o mini tts and ElevenLabs v2 π₯ > Trained on 10M hours (speech, music, events) > Built on top of Llama 3.2 3B > Works real-time and on edge > Beats GPT-4o-mini-tts, ElevenLabs v2 in prosodyβ¦
This is amazing.
Qwen on a rampage π₯ - Qwen3 Coder 480B (35B Active), beats Kimi K2 AND Claude Sonnet 4 - Apache 2.0 licensed! > Upto 1M context > Non-Reasoning > Supports agentic and browser mode So so excited for the smaller models in the series π€
won't finish it tonight but will be worth the wait, I just got some very fun results :)
I think tonight I'll have time to nail a little tutorial on: Automatic System Prompt Optimization. I hope some of you are up for that :)
I think tonight I'll have time to nail a little tutorial on: Automatic System Prompt Optimization. I hope some of you are up for that :)
Dspy is a slippery slope. First, you just want a better prompt. Next thing you know, you want to fine tune a smaller model.