Stephen with a ph...D
@stephenmcgeephd.bsky.social
Data nerd focused on integrating AI in into healthcare genetics
Fascinating (albeit ominously titled) dataset for benchmarking emerging multi-modal models - Humanity’s Last Exam
Over 3000 questions gathered from a global collaborative of over 1000 contributors from >50 countries.
lastexam.ai
Over 3000 questions gathered from a global collaborative of over 1000 contributors from >50 countries.
lastexam.ai
Humanity's Last Exam
Humanity's Last Exam Dataset
lastexam.ai
February 2, 2025 at 2:26 PM
Fascinating (albeit ominously titled) dataset for benchmarking emerging multi-modal models - Humanity’s Last Exam
Over 3000 questions gathered from a global collaborative of over 1000 contributors from >50 countries.
lastexam.ai
Over 3000 questions gathered from a global collaborative of over 1000 contributors from >50 countries.
lastexam.ai
Microsoft pushed out a smarter retrieval paradigm, Chain-of-thought RAG (CoRAG) - how to leverage o1-like models to retrieve and reason to better refine queries at each step thereby improving accuracy and helping to escape limitations of conventional RAG systems #MLSky - arxiv.org/abs/2501.143...
Chain-of-Retrieval Augmented Generation
This paper introduces an approach for training o1-like RAG models that retrieve and reason over relevant information step by step before generating the final answer. Conventional RAG methods usually p...
arxiv.org
January 29, 2025 at 12:28 PM
Microsoft pushed out a smarter retrieval paradigm, Chain-of-thought RAG (CoRAG) - how to leverage o1-like models to retrieve and reason to better refine queries at each step thereby improving accuracy and helping to escape limitations of conventional RAG systems #MLSky - arxiv.org/abs/2501.143...
Just realized that with images, OpenAI inflates prompt tokens for billing purposes when using GPT-4o-mini to equal image costs for GPT-4o. Not a deal-breaker, by any means, but the cost-efficient promise of 4o-mini seems to be text-specific.
January 29, 2025 at 5:04 AM
Just realized that with images, OpenAI inflates prompt tokens for billing purposes when using GPT-4o-mini to equal image costs for GPT-4o. Not a deal-breaker, by any means, but the cost-efficient promise of 4o-mini seems to be text-specific.
Recently saw a post stating OCR is dead. I don’t think we’re quite there yet. Though multimodal models have dramatically improved structuring clinical docs, OCR still has much to offer that VLMs haven’t yet conquered (bounding box detection, hallucinations in poor quality images, etc)
#MLSky
#MLSky
January 25, 2025 at 2:19 PM
Recently saw a post stating OCR is dead. I don’t think we’re quite there yet. Though multimodal models have dramatically improved structuring clinical docs, OCR still has much to offer that VLMs haven’t yet conquered (bounding box detection, hallucinations in poor quality images, etc)
#MLSky
#MLSky
Reposted by Stephen with a ph...D
The RL book by Kevin Murphy is finally online (copied shamelessly from the other place) arxiv.org/abs/2412.05265
Reinforcement Learning: An Overview
This manuscript gives a big-picture, up-to-date overview of the field of (deep) reinforcement learning and sequential decision making, covering value-based RL, policy-gradient methods, model-based met...
arxiv.org
December 9, 2024 at 6:25 AM
The RL book by Kevin Murphy is finally online (copied shamelessly from the other place) arxiv.org/abs/2412.05265
I’ve been really digging this approach to GraphRAG lately by Circlemind - PageRank FTW!!! - github.com/circlemind-a...
#AI #machinelearning
#AI #machinelearning
GitHub - circlemind-ai/fast-graphrag: RAG that intelligently adapts to your use case, data, and queries
RAG that intelligently adapts to your use case, data, and queries - circlemind-ai/fast-graphrag
github.com
November 23, 2024 at 11:28 AM
I’ve been really digging this approach to GraphRAG lately by Circlemind - PageRank FTW!!! - github.com/circlemind-a...
#AI #machinelearning
#AI #machinelearning