🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

👉 Our #NeurIPS2024 paper on Chain of Thoughtlessness @ 11AM poster session today (East Hall #3010, 11AM-2pm...). All three of us are here and looking forward to chat/answer qns.. 🙏

With some Yochanites--past and present--at #NeurIPS2024 🤗 (L-R: kstechly Karthik Valmeekam Subbarao Kambhampati (కంభంపాటి సుబ్బారావు) Sarath Sreedharan Kartik Talamadupula )

🥳🎉Congrats to Yochanite Karthik Valmeekam Karthik Valmeekam on being selected as an IBM PhD Fellow! Thanks IBM Research for this recognition of his ongoing work on the planning/reasoning abilities of LLMs!

Congrats to Karthik Valmeekam, a #PhD student working under the supervision of Subbarao Kambhampati (కంభంపాటి సుబ్బారావు) in #SCAI, who has received an IBM PhD Fellowship Award from IBM Research. Given since 1951, the award recognizes research excellence in PhD students addressing areas of great technological interest.

📢 DeepSeek-R1 on PlanBench 📢 DeepSeek-R1 gets similar performance as OpenAI’s o1 (preview)—achieving 96.6% on Blocksworld and 39.8% on its obfuscated version, Mystery BW. The best part? ⚡It’s 21x cheaper than o1-preview, offering similar results at a fraction of the cost!

📢"On the Self-Verification Limitations of LLMs in Reasoning and Planning Tasks" arxiv.org/abs/2402.08115 with kstechly and Karthik Valmeekam apparently made it to #ICLR2025.. Swimming🏊 to Singapore.. 😎 [The 🧵s below give the details.. ]

On the use of Verifiers with LLMs--External vs. Internal LLM-Modulo (with kstechly, Karthik Valmeekam and @21st_Warlock ) We are a bit tickled that verifiers seem to be all the rage on the AI twitter, as we have been advocating use of external verifiers of various

On the MDP formulation of LLMs used in R1 [Not quite #SundayHarangue] Everyone knows that R1 is using RL on LLMs. Most also know that RL is done on an underlying MDP formulation. Not everyone might have grokked the rather strange nature of the MDP formulation used. We had a fun

Back in September, when we evaluated o1, and used LRM (Large Reasoning Model) to refer to that type of models, there was some pushback. Now looking at many arXiv ICML submissions analyzing these beasts, I see that LRM is becoming pretty standard usage. That's so fetch.. 😋

[1/3] I am incredibly humbled and honored to be selected as one of IEEE AI's 10 to watch. However, this wouldn't have been possible without my wonderful collaborators. To start with, I will always be thankful that Subbarao Kambhampati (కంభంపాటి సుబ్బారావు) agreed to work with me.

📢Delighted to share that our analysis of the planning abilities of 🍓 o1 has now been accepted to #TMLR Transactions on Machine Learning Research . Joint work with Karthik Valmeekam kstechly and Atharva Gundawar. Final version, including DeepSeek R1 results, appearing soon..

📢 Tomorrow (3rd April) Karthik Valmeekam Karthik Valmeekam will will try and hoodwink his PhD committee (comprising Thomas G. Dietterich Huan Liu Siddharth Srivastava and Sarath Sreedharan) to allow him to advance to PhD candidacy.. 🤞

Our invited commentary for the Annals of NYAS titled "(How) Do reasoning models reason?" is now online 👉 nyaspubs.onlinelibrary.wiley.com/doi/epdf/10.11… It is a written version of my recent talks (and #SundayHarangues) on the recent developments in LRMs.. (w/ kstechly & Karthik Valmeekam )

PSA for #ICLR2025 authors frantically making posters: Stop worrying! Just prompt o3 and GPT-4o and you will have an AGI-ready poster in seconds! Here is one kstechly and Karthik Valmeekam cooked up--and it looks fully legit from poster distance!

Enroute to #ICLR2025 with kstechly .. and itching for a #ThursdayHarangue..

It ain't "The Bitter Lesson" if you are in the loop curating the training data for your LLM, y'all.. Pick your lesson, will ya? #SundayHarangue (h/t kstechly)

Yochanites at one of those rare group breakfasts! (after that 5AM #NeurIPS2025 deadline..)

So kstechly 🎓 from Arizona State University yesterday. Kaya was a rather unusual Yochanite-- didn't take any courses or do theses with me; and wasn't even in a CS degree! She was ever present in lab, and lunches though--and a force in all group meetings and many papers. She will be missed..

Some of what that recent Apple LRM limitations paper shows is known (pardon my friendly Schmidhubering; I do welcome more LLM studies with scientific skepticism). Our study 👇 from Sep 2024 shows o1 accuracy degrading as complexity increases.. 1/ x.com/rao2z/status/1…

Everyone tweetin about the Apple paper, hope you read Subbarao Kambhampati (కంభంపాటి సుబ్బారావు) on reasoning models