Harsha
@harshanelaturu.bsky.social
Applied Scientist Intern @ Amazon | Community lead @ Cohere For AI | Prev MPI-INF, CISPA, Media Lab.
I work on systems and algorithms for efficient training/inference of large scale machine learning models.
I work on systems and algorithms for efficient training/inference of large scale machine learning models.
Reposted by Harsha
Can we predict which reasoning paths will succeed before seeing the answer? 🤔
Our new paper (arxiv.org/abs/2510.10494) proposes latent-trajectory signals from LLMs' hidden states to identify high-quality reasoning, cutting inference costs by up to 70% while maintaining accuracy
Our new paper (arxiv.org/abs/2510.10494) proposes latent-trajectory signals from LLMs' hidden states to identify high-quality reasoning, cutting inference costs by up to 70% while maintaining accuracy
Tracing the Traces: Latent Temporal Signals for Efficient and Accurate Reasoning
Reasoning models improve their problem-solving ability through inference-time scaling, allocating more compute via longer token budgets. Identifying which reasoning traces are likely to succeed remain...
arxiv.org
October 22, 2025 at 3:38 PM
Can we predict which reasoning paths will succeed before seeing the answer? 🤔
Our new paper (arxiv.org/abs/2510.10494) proposes latent-trajectory signals from LLMs' hidden states to identify high-quality reasoning, cutting inference costs by up to 70% while maintaining accuracy
Our new paper (arxiv.org/abs/2510.10494) proposes latent-trajectory signals from LLMs' hidden states to identify high-quality reasoning, cutting inference costs by up to 70% while maintaining accuracy
Excited to host Andrei, PhD student @ IST who will present 'Pushing the limits of LLM quantization via the linearity theorem' on Jan 10 @ 1800 CET at Cohere For AI. Really cool results and look forward to the talk. Join the community: tinyurl.com/C4AICommunit...
January 4, 2025 at 9:44 PM
Excited to host Andrei, PhD student @ IST who will present 'Pushing the limits of LLM quantization via the linearity theorem' on Jan 10 @ 1800 CET at Cohere For AI. Really cool results and look forward to the talk. Join the community: tinyurl.com/C4AICommunit...
I was a beta-tester on #ghostty [though I got in towards the end of the period]. I primarily used alacritty before -- and both were very fast and I would say on average I might not be able to tell them apart. However, I love the customizability, snappiness and well-thought out it is.
December 28, 2024 at 5:57 PM
I was a beta-tester on #ghostty [though I got in towards the end of the period]. I primarily used alacritty before -- and both were very fast and I would say on average I might not be able to tell them apart. However, I love the customizability, snappiness and well-thought out it is.
Reposted by Harsha
Yes, and the original sin of computer science was its haughty conception as a subfield of mathematics rather than an engineering discipline.
November 28, 2024 at 4:03 PM
Yes, and the original sin of computer science was its haughty conception as a subfield of mathematics rather than an engineering discipline.
Time to find all the ML Systems and Efficiency folks on Bluesky :)
November 24, 2024 at 10:43 AM
Time to find all the ML Systems and Efficiency folks on Bluesky :)