The GPQA (Graduate-Level Google-Proof Q&A) benchmark is a challenging dataset designed to assess the capabilities of large language models (LLMs) in handling complex, domain-specific questions in biology, physics, and chemistry. It consists of 448 multiple-choice questions
@testingcatalog
-
o1 reaches level 2 for writing tasks
By
–
o1 has a chance to mark level 2 for writing tasks finally It feels to have much better instruction-following than all previous models. Looking forward to testing it with web search capabilities. Such a model, paired with a big RAG would also make it huge for research!
-
O1 Model File Upload Access Removed
By
–
o1 next week? The o1 model (non-preview?) WAS accessible with file upload support! Now it seems to be patched. Pointing to o1 in the query directly allowed users to upload files and trigger the "thinking" process.
-
Claude 3.5 Sonnet to Preview Visual PDFs
By
–
BREAKING : Claude is getting a new feature preview Visual PDFs! Friday feature drop "Give Claude 3.5 Sonnet the ability to view and analyze images, charts, and graphs in PDFs, in addition to text. PDFs that are <100 pages are supported."
-
New Era of AI-Powered Search in 2024
By
–
90% accuracy or so? 👀👀👀
— 🚨 AI News | TestingCatalog (@testingcatalog) 31 octobre 2024
The new era of search started in October 2024. Today we’ve got most of pieces of the puzzle.
– Claude desktop app
– SearchGPT
– AI Studio search grounding
Perplexity has been a trigger of the shift.
Looking forward to play with Pandragon next 👀 https://t.co/VQKvBzP8tq90% accuracy or so? The new era of search started in October 2024. Today we’ve got most of pieces of the puzzle. – Claude desktop app
– SearchGPT – AI Studio search grounding Perplexity has been a trigger of the shift. Looking forward to play with Pandragon next -
Google AI Studio Adds Search Grounding Feature
By
–
BREAKING : Google AI Studio got Search Grounding. Now you can play with web search directly on AI Studio as well as use grounding adjustment!
-
ChatGPT generates first AVM voice clip
By
–
Welcome a first AVM voice clip made via ChatGPT app! 🔥
— 🚨 AI News | TestingCatalog (@testingcatalog) 30 octobre 2024
This feature is still in development but now it can create an actual file which I was able to download. It is capturing past 25 seconds of a voice conversation into a video file (example is trimmed) pic.twitter.com/MTxd5aN4UMWelcome a first AVM voice clip made via ChatGPT app! This feature is still in development but now it can create an actual file which I was able to download. It is capturing past 25 seconds of a voice conversation into a video file (example is trimmed)
-
ChatGPT Slack and Google Workspace integrations revealed
By
–
ChatGPT is working on Slack and Google Workspace integration for enterprise customers. As discovered by Tibor, ChatGPT may get 2 new connectors and a separate model "gpt-4o-gdrive" Slack integration will potentially let LLM access public Slack channels.
-
GitHub Copilot Adds Claude 3.5 and Gemini 1.5 Pro
By
–
Github Copilot got support for Claude 3.5 (new) and Gemini 1.5 Pro models besides previously available models from OpenAI.
— 🚨 AI News | TestingCatalog (@testingcatalog) 29 octobre 2024
A collab you would not expect ❤️ https://t.co/OEsaiefUNX pic.twitter.com/E1SqCST6C2Github Copilot got support for Claude 3.5 (new) and Gemini 1.5 Pro models besides previously available models from OpenAI. A collab you would not expect
-
Google Adds Custom Prompts to NotebookLM
By
–
WIP : Google is working on the possibility of setting system prompts to Notebooks on NotebookLM Users will be able to specify a "custom prompt" and a style guide. Current placeholders suggest that it may be a way to specify a "persona" of the "guest"?