Distill Whisper uses knowledge distillation to train a smaller model to achieve 99% accuracy of a larger one with 2% data, aiming for real-time AI voice applications without delays.
GENERATIVE AI
-
GPT-4.5 Turbo released in secret?
By
–
Was GPT-4.5 Turbo released in stealth? Numerous ChatGPT Plus users are getting this response. Mass hallucination or something is cooking.
-
Quip: 2-Bit Quantization for Large Language Models
By
–
10/ Quip – compresses trained model weights into a lower precision format; combines lattice codebooks with incoherence processing to create 2 bit quantized models; significantly closes the gap between 2 bit quantized LLMs and unquantized 16 bit models.
-
LLMs in Medicine: Comprehensive Survey of Applications and Challenges
By
–
6/ LLMs in Medicine – a comprehensive survey (analyzing 300+ papers) on LLMs in medicine; includes an overview of the principles, applications, and challenges faced by LLMs in medicine.
-
Weak-to-Strong Generalization: Eliciting Full Capabilities of Strong Models
By
–
2/ Weak-to-strong Generalization – studies if weak model supervision can elicit the full capabilities of stronger models; when naively fine-tuning strong pretrained models on weak model generated labels they can perform better than their weak supervisors.
-
Audiobox: Unified Flow-Matching Model for Audio Generation
By
–
3/ Audiobox – a unified model based on flow-matching capable of generating various audio modalities; designs description-based and example-based prompting to enhance controllability and unify speech and sound generation paradigms.https://t.co/OcXaDuRU6j
— DAIR.AI (@dair_ai) 17 décembre 20233/ Audiobox – a unified model based on flow-matching capable of generating various audio modalities; designs description-based and example-based prompting to enhance controllability and unify speech and sound generation paradigms.
-
Top ML Papers of the Week: SLAM, LLMs, Medical AI
By
–
The Top ML Papers of the Week (Dec 11 – Dec 17): – Gaussian-SLAM
– LLMs in Medicine
– Mathematical LLMs
– Beyond Human Data for LLMs
– Weak-to-strong Generalization
– Towards Fully Transparent Open-Source LLM
… -
Reddit User Reports Significant Coding Advantage Over GPT-4 API
By
–
Tbf Reddit user felt a huge advantage in coding compared to gpt4.0 api.
-
GPT 4.5 Release Boosts Coding Skills for Developers
By
–
Several people on Reddit realize huge increases in in coding skills via GPT 4.5. seems to be a real release
-
GPT 4.5 Turbo Confirmed According to Multiple Reddit Users
By
–
GPT 4.5 Turbo confirmed. Several users on Reddit confirm this