AI300

Tri Dao / Chief Scientist at Together
3721.
Clarification: I was comparing A @ B + C here, where the cute-dsl version is quite good at overlapping the epilogue.›
LLM score 85 · about 1 year ago
Sheryl Hsu / OpenAI Researcher
3722.
We officially entered the 2025 International Olympiad in Informatics (IOI) online competition track and adhered to the same restrictions as the human contestants, including submissions and time limits, but without direct supervision from the contest organizers.
LLM score 70 · about 1 year ago
Alexander Wei / OpenAI Researcher
3723.
3/ We’ve come a long way since last summer.›
LLM score 70 · about 1 year ago
Sheryl Hsu / OpenAI Researcher
3724.
5/n It’s been really exciting to see the progress of our newest research methods at OpenAI, with our successes at the AtCoder World Finals, IMO, and IOI over the last couple weeks.›
LLM score 20 · about 1 year ago
Sheryl Hsu / OpenAI Researcher
3725.
4/n This result demonstrates a huge improvement over @OpenAI’s attempt at IOI last year where we finished just shy of a bronze medal with a significantly more handcrafted test-time strategy.›
LLM score 80 · about 1 year ago
Juntang Zhuang / MTS at xAI (pre-training lead)
3726.
It’s extremely fun though tough to train the first natively multimodal model ever in xAI.›
LLM score 92 · about 1 year ago
Geoffrey Hinton
3727.
A major cut to the funding of the National Science Foundation would be very bad for the future of the US.›
LLM score 20 · about 1 year ago
Ted Sanders / OpenAI Researcher
3728.
a cool thing you get to see building AI products: ›
LLM score 75 · about 1 year ago
Ted Sanders / OpenAI Researcher
3729.
GPT-5 is here! it's way better at coding - not just in pointless evals, but real usage.›
LLM score 70 · about 1 year ago
Jeremy Bernstein / Thinking Machines Researcher
3730.
I had wondered why there was no official Dion implementation by the authors...›
LLM score 75 · about 1 year ago
Sally Zhu / Researcher at Flapping Airplanes
3731.
Interesting that you need the teacher + student to share the same base model; reminds me of linear mode connectivity stuff.›
LLM score 85 · over 1 year ago
Tri Dao / Chief Scientist at Together
3732.
Hierarchical layout is super elegant.›
LLM score 85 · over 1 year ago
Jason Wei / AI Researcher at Meta
3733.
Becoming an RL diehard in the past year and thinking about RL for most of my waking hours inadvertently taught me an important lesson about how to live my own life.›
LLM score 60 · over 1 year ago
Jason Wei / AI Researcher at Meta
3734.
New blog post about asymmetry of verification and "verifier's law": https://t.co/bvS8HrX1jP›
LLM score 80 · over 1 year ago
Jakub Pachocki / OpenAI Chief Scientist
3735.
I am extremely excited about the potential of chain-of-thought faithfulness & interpretability.›
LLM score 80 · over 1 year ago
Lilian Weng / Thinking Machines Cofounder
3736.
I still find it mysterious whether and how intelligence and capabilities transfer between domains and skills - from meta learning during early days to more recent question like whether solving maths helps writing a good essay.›
LLM score 75 · over 1 year ago
Tri Dao / Chief Scientist at Together
3737.
I played w it for 1h. Went through my usual prompts (math derivations, floating point optimizations, …).›
LLM score 35 · over 1 year ago
Tri Dao / Chief Scientist at Together
3738.
@RaghuGanti @cHHillee Oh you’d want to use warp reduction if the whole row fits into 1 warp.›
LLM score 80 · over 1 year ago
Tri Dao / Chief Scientist at Together
3739.
They’ve finally done it. They got rid of tokenizers! https://t.co/x4CXHdCw0W
LLM score 60 · over 1 year ago
Tri Dao / Chief Scientist at Together
3740.
I really like Phil Tillet's framing of different tools having different tradeoffs in productivity and performance: torch compile, triton, CUDA, PTX.›
LLM score 80 · over 1 year ago
Tri Dao / Chief Scientist at Together
3741.
Getting mem-bound kernels to speed-of-light isn't a dark art, it's just about getting the a couple of details right.›
LLM score 85 · over 1 year ago
Tri Dao / Chief Scientist at Together
3742.
Albert articulates really well the trade offs between transformers and SSMs.›
LLM score 80 · over 1 year ago
Tri Dao / Chief Scientist at Together
3743.
Turns out you can do length generalization for recurrent model by simply training for another extra 100 steps with a careful choice of initial states https://t.co/QgyrFBSFQh
LLM score 80 · over 1 year ago
Shuchao Bi / Meta Researcher
3744.
If you are curious about how to pass speech Turing test, @pika7ma and I shared some high level learnings here: https://t.co/0ujh1i1aK6
LLM score 65 · over 1 year ago
Yang Chen / Nvidia Research Scientist
3745.
The first thing we did was to make sure the eval setup is correct!›
LLM score 92 · over 1 year ago
Yang Chen / Nvidia Research Scientist
3746.
📢We conduct a systematic study to demystify the synergy between SFT and RL for reasoning models.›
LLM score 92 · over 1 year ago
Geoffrey Hinton
3747.
Congratulations to @Yoshua_Bengio on launching @LawZero_ — a research effort to advance safe-by-design AI, especially as frontier systems begin to exhibit signs of self-preservation and deceptive behaviour.
LLM score 75 · over 1 year ago
Yang Chen / Nvidia Research Scientist
3748.
Does RL incentive reasoning capability over the starting SFT model? ›
LLM score 92 · over 1 year ago
Geoffrey Hinton
3749.
I just watched a great compilation of various people's views about what is coming:›
LLM score 10 · over 1 year ago
Ludwig Schmidt / Anthropic MTS
3750.
Very excited about our new agent benchmark! I think it's a nice way of evaluating how well agents can do complex task in terminal (command line) environments.›
LLM score 65 · over 1 year ago