Technically it supports fewer languages than whisper, 40 vs 99
The main problem isn't "bother", it's training data. You need hundreds of thousands of hours of high quality transcripts to train models like these and that just doesn't exist for like zulu or whatever
This report introduces Dolphin, a large-scale multilingual automatic speech recognition (ASR) model that extends the Whisper architecture to support a wider range of languages. Our approach integrates in-house proprietary and open-source datasets to refine and optimize Dolphin's performance. The mod...
I want to clarify something. Reranker is a general term that can refer to any model used for reranking. It is independent of implementation.
What you refer to
because reranker models look at the two pieces of content simultaneously and can be fine tuned to the domain in question. They shouldn't be used for the initial retrieval because the evaluation time is O(n²) as each combination of input
Is a specific implementation known as CrossEncoder that is common for reranking models but not retrieval ones for the reasons you described. But you can also use any other architecture
Link to bluesky https://bsky.app/profile/tomaarsen.com/post/3llc2jvwah22f
Some more details https://huggingface.co/blog/train-reranker
Claude frequently draws svgs to illustrate things for me (I'm guessing it's in the prompt) but even though it's better at it than all the other models, it still kinda sucks. It's just fudamentally dumb task to do for a purely language model, similar to the arc-agi benchmark , just makes more sense for a vision model and trying to get an llm to do is a waste
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
what is the license? The link on hf just 404s
Very similar to chain of draft but seems more thorough
Recent advances in large language models have demonstrated remarkable reasoning capabilities through Chain of Thought (CoT) prompting, but often at the cost of excessive verbosity in their intermediate outputs, which increases computational overhead. We introduce Sketch-of-Thought (SoT), a novel pro...
It matches R1 in the given benchmarks. R1 has 671B params (36 activated) while this only has 32
Large Language Models (LLMs) have demonstrated remarkable performance in solving complex reasoning tasks through mechanisms like Chain-of-Thought (CoT) prompting, which emphasizes verbose, step-by-step reasoning. However, humans typically employ a more efficient strategy: drafting concise intermedia...
Atom of Thoughts (AOT): lifts gpt-4o-mini to 80.6% F1 on HotpotQA, surpassing o3-mini and DeepSeek-R1 ! For each reasoning step, it: 1. Decompose the question into DAG 2. Contract the subquestions into a NEW simpler question 3. Iterate until reaching an atomic question
