"Transcendence" is when an LLM, trained on diverse data from many experts, can exceed the ability of the individuals in its training data.
This paper demonstrates three types: when AI picks the right expert skill to use, when AI has less bias than experts & when it generalizes.
📚 Read
@datascienceiot
This paper demonstrates three types: when AI picks the right expert skill to use, when AI has less bias than experts & when it generalizes.
📚 Read
@datascienceiot
Tencent AI Lab introduces Vision-SR1
A self-rewarding Vision-Language Model to fix visual hallucinations & language shortcuts
📚Paper: https://huggingface.co/papers/2508.19652
📚Code: https://github.com/zli12321/Vision-SR1.
@datascienceiot
A self-rewarding Vision-Language Model to fix visual hallucinations & language shortcuts
📚Paper: https://huggingface.co/papers/2508.19652
📚Code: https://github.com/zli12321/Vision-SR1.
@datascienceiot
"GPT-5 moves from human-comparable to above human-expert performance"
GPT-5 outperforms licensed human experts by 25-30% and achieves SOTA results on the US medical licensing exam and the MedQA benchmark.
I sound like a broken record, but AI models are better than most doctors.
📚 Paper
@datascienceiot
GPT-5 outperforms licensed human experts by 25-30% and achieves SOTA results on the US medical licensing exam and the MedQA benchmark.
I sound like a broken record, but AI models are better than most doctors.
📚 Paper
@datascienceiot
On the Theoretical Limitations of Embedding-Based Retrieval"
📚Статья: https://arxiv.org/abs/2508.21038
@datascienceiot
📚Статья: https://arxiv.org/abs/2508.21038
@datascienceiot
Forwarded from Machinelearning
400 страниц про всё, что нужно знать об агентных системах. Автор — senior engineer в Google, выложил драфт для открытого ревью.
📖 В книге:
- продвинутые техники промптинга
- паттерны для мульти-агентов
- использование инструментов и MCP
- практические примеры с кодом
⚡ По сути, это полный справочник по построению умных агентов. Must-read для разработчиков AI.
@ai_machinelearning_big_data
#AI #Agents #Google #OpenSource #freebook
Please open Telegram to view this post
VIEW IN TELEGRAM
Please open Telegram to view this post
VIEW IN TELEGRAM
Please open Telegram to view this post
VIEW IN TELEGRAM
Please open Telegram to view this post
VIEW IN TELEGRAM
Please open Telegram to view this post
VIEW IN TELEGRAM
Please open Telegram to view this post
VIEW IN TELEGRAM
SFR-DeepResearch: Towards Effective Reinforcement Learning for Autonomously Reasoning Single Agents
📚 Читать
@datascienceiot
@datascienceiot
Please open Telegram to view this post
VIEW IN TELEGRAM
Aligning LLMs for the Classroom with Knowledge-Based Retrieval -- A Comparative RAG Study
📚 Читать
@datascienceiot
@datascienceiot
Please open Telegram to view this post
VIEW IN TELEGRAM