@ikyle.bsky.social
Reposted
100,277 tokens – The vocabulary size of GPT-4’s tokenizer, showcasing the complexity of tokenization.
$40,000 (2019) → $100 (2024) – The reduction in cost to train a GPT-2 equivalent model, illustrating improvements in efficiency and hardware advancements.
$40,000 (2019) → $100 (2024) – The reduction in cost to train a GPT-2 equivalent model, illustrating improvements in efficiency and hardware advancements.
February 6, 2025 at 2:28 AM
100,277 tokens – The vocabulary size of GPT-4’s tokenizer, showcasing the complexity of tokenization.
$40,000 (2019) → $100 (2024) – The reduction in cost to train a GPT-2 equivalent model, illustrating improvements in efficiency and hardware advancements.
$40,000 (2019) → $100 (2024) – The reduction in cost to train a GPT-2 equivalent model, illustrating improvements in efficiency and hardware advancements.
Reposted
Paper: Large Language Models Share Representations of Latent Grammatical Concepts Across Typologically Diverse Languages by one of the author - @jannikbrinkmann.bsky.social ( arxiv.org/abs/2501.06346 )
Repo: github.com/jannik-brink...
Post on X: x.com/jannikbrinkm...
Repo: github.com/jannik-brink...
Post on X: x.com/jannikbrinkm...
Large Language Models Share Representations of Latent Grammatical Concepts Across Typologically Diverse Languages
Human bilinguals often use similar brain regions to process multiple languages, depending on when they learned their second language and their proficiency. In large language models (LLMs), how are mul...
arxiv.org
February 6, 2025 at 7:19 AM
Paper: Large Language Models Share Representations of Latent Grammatical Concepts Across Typologically Diverse Languages by one of the author - @jannikbrinkmann.bsky.social ( arxiv.org/abs/2501.06346 )
Repo: github.com/jannik-brink...
Post on X: x.com/jannikbrinkm...
Repo: github.com/jannik-brink...
Post on X: x.com/jannikbrinkm...
Reposted
Llama2’s internal “lingua franca” is not English, but concepts - and, crucially, concepts that are biased toward English. Hence, English could still be seen as an “internal language”, but in a semantic, rather than a purely lexical.
Paper: arxiv.org/abs/2402.10588
Post on X: x.com/cervisiarius...
Paper: arxiv.org/abs/2402.10588
Post on X: x.com/cervisiarius...
February 6, 2025 at 7:26 AM
Llama2’s internal “lingua franca” is not English, but concepts - and, crucially, concepts that are biased toward English. Hence, English could still be seen as an “internal language”, but in a semantic, rather than a purely lexical.
Paper: arxiv.org/abs/2402.10588
Post on X: x.com/cervisiarius...
Paper: arxiv.org/abs/2402.10588
Post on X: x.com/cervisiarius...
Reposted
Paper: arxiv.org/abs/2502.03382
Samples: hf.co/spaces/kyuta...
Inference code: github.com/kyutai-labs/...
Models: huggingface.co/kyutai
Post on X: x.com/kyutai_labs/...
Samples: hf.co/spaces/kyuta...
Inference code: github.com/kyutai-labs/...
Models: huggingface.co/kyutai
Post on X: x.com/kyutai_labs/...
High-Fidelity Simultaneous Speech-To-Speech Translation
We introduce Hibiki, a decoder-only model for simultaneous speech translation. Hibiki leverages a multistream language model to synchronously process source and target speech, and jointly produces tex...
arxiv.org
February 7, 2025 at 7:41 AM
Paper: arxiv.org/abs/2502.03382
Samples: hf.co/spaces/kyuta...
Inference code: github.com/kyutai-labs/...
Models: huggingface.co/kyutai
Post on X: x.com/kyutai_labs/...
Samples: hf.co/spaces/kyuta...
Inference code: github.com/kyutai-labs/...
Models: huggingface.co/kyutai
Post on X: x.com/kyutai_labs/...
Reposted
Demo: playground.zyphra.com/audio
Blog: zyphra.com/post/beta-re...
Models:
huggingface.co/Zyphra/Zonos... and huggingface.co/Zyphra/Zonos...
Repo: github.com/Zyphra/Zonos
Blog: zyphra.com/post/beta-re...
Models:
huggingface.co/Zyphra/Zonos... and huggingface.co/Zyphra/Zonos...
Repo: github.com/Zyphra/Zonos
Zyphra Maia
playground.zyphra.com
February 10, 2025 at 6:44 PM