Subbarao Kambhampati (కంభంపాటి సుబ్బారావు) ✓
@rao2z · Tempe, AZ · joined 30 Oct 2014
AI researcher & teacher @SCAI_ASU. Former President of @RealAAAI; Chair of @AAAS Sec T. Here to tweach #AI. YouTube Ch: https://t.co/4beUPOmf6y Bsky: rao2z
28 263Followers
67Following
11 990Posts total
387.1KViews on collected posts
Latest posts
Looking for new posts…
@rao2z I've said that for months, every time some doomer talking point comes up. If the Chinese are distilling your models, pause for two months. If the Chinese subsequently stall, you're right. If not, you've got two months' loot to put on the S-1.
2.6K views · 50 likes
· 0 reposts · 1 replies
Open on X →
..and like I said, there is *always* a relevant Seinfeld reference..
https://t.co/DGdBamRnoZ
5.6K views · 8 likes
· 0 reposts · 0 replies
Open on X →
I don't quite understand why Anthropic is worried about about China in implementing their slow down. If they are right and China got here only by distilling Ant's models, then Ant slowing down should slow China too, no? 🤔
33.4K views · 695 likes
· 36 reposts · 35 replies
Open on X →
Besides, agentic evaluation has increasingly little to do with NLP, and is really a planning err.. AI-Complete problem..🙃
3K views · 13 likes
· 2 reposts · 0 replies
Open on X →
The New York AI Times.. https://t.co/BXgPNBZ7TA
1.4K views · 10 likes
· 2 reposts · 1 replies
Open on X →
We all can agree @stanfordnlp is far ahead of most other academic labs when it comes to modern frontier AI. However, a fair question to ask is how can we entrust Stanford NLP as this sole arbiter role when most of its students and faculty have deep financial entanglements with
16.8K views · 152 likes
· 15 reposts · 11 replies
Open on X →
If your agents escaped your sandbox, may be its because you are lousy at building sandboxes--and not necessarily because the agents are conniving super-intelligent entities.. 🤔
20.1K views · 95 likes
· 24 reposts · 8 replies
Open on X →
In a recent harrowing development, a very normal routine experiment by the frontier company Ant went completely off the rails. 😱
Ant put thousands of their ant agents in a pretty secure ant farm with mesh walls and all, exhorted them to not to get out of the secure ant farm, htt
29.5K views · 100 likes
· 21 reposts · 8 replies
Open on X →
@rao2z “The ability to discern patterns in mountains of data so as to be able to predict outcomes, or to generate new patterns based on the past, isn’t an exercise in intelligence, & the anthropomorphism employed here affects the way we think about these issues.”
https://t.co/wFD
1.3K views · 5 likes
· 1 reposts · 0 replies
Open on X →
Here is the link to the paper, if you are interested 👉https://t.co/6b9hP9ASh9 7/ https://t.co/IspEMQsDDf
2.8K views · 30 likes
· 2 reposts · 2 replies
Open on X →
Some of these arguments may be familiar for those who follow my #SundayHarangues, as these have been cooking for a while.. 🤷♂️ 7/
2.5K views · 8 likes
· 0 reposts · 1 replies
Open on X →
..and finally the analysis of RL post-training on intermediate token length and interpretability claims.. 👇 5/
https://t.co/GWMbJJQP8J
2.4K views · 11 likes
· 0 reposts · 1 replies
Open on X →
In addition to pointing out the pitfalls of the intermediate token anthropomorphization, the paper also suggests some alternate views of the effectiveness of LRMs that don't require anthropomorphization. 6/
1.6K views · 12 likes
· 0 reposts · 1 replies
Open on X →
..and a parallel effort on the paradoxes of intermediate token interpretability in Q&A tasks 👇 4/
https://t.co/pPs1LRVNwa
2.7K views · 15 likes
· 0 reposts · 1 replies
Open on X →
We support the position with emerging work both from our and other groups. Supporting efforts from our group include the Beyond Semantics paper that shows a disconnect between intermediate token validity and solution accuracy in planning tasks 👇3/
https://t.co/jnci6lfiP6
3.6K views · 18 likes
· 0 reposts · 1 replies
Open on X →
These anthropomorphization tendencies include both viewing intermediate tokens as interpretable traces of LLM's "thinking" and confusing the length of the intermediate tokens as indicative of the "thinking effort" 2/
3K views · 34 likes
· 1 reposts · 1 replies
Open on X →
Anthropomorphization of intermediate tokens as reasoning/thinking traces isn't quite a harmless fad, and may be pushing LRM research into questionable directions.. So we decided to put together a more complete argument.. 👇🧵 1/ https://t.co/x6aElcdAq4
108.1K views · 486 likes
· 86 reposts · 12 replies
Open on X →
This RLiNo? paper lead by @soumya_samineni & @durgesh_kalwar dives into the MDP model used in the RL post-training methods inspired by DeepSeek R1, and asks if some of the idiosyncrasies of RL in R1 aren't just consequences of the simplistic structural assumptions in the MDP 🧵1/
38.9K views · 54 likes
· 12 reposts · 4 replies
Open on X →
Semantics of Intermediate Tokens in Trace-based distillation in Q&A tasks: Yochanites @sbhambr1 and @biswas_2707 looked at distillation on a Q&A task, and found a disconnect between the validity of derivational traces and the correctness of the solution.. 🧵 1/ https://t.co/3jpt2Y
10.7K views · 27 likes
· 8 reposts · 2 replies
Open on X →
Do Intermediate Tokens Produced by LRMs (need to) have any semantics? Our new study "Beyond Semantics: The Unreasonable Effectiveness of Reasonless Intermediate Tokens" lead by @kayastechly, @karthikv792 @_gundawar & @PalodVardh12428 dives into this question 🧵 1/ https://t.co/9S
96.9K views · 400 likes
· 68 reposts · 16 replies
Open on X →
Against accounts of the same size
8 posts from the last 90 days, next to the 10K–100K follower range. shown widely, but few of those viewers react.
Median views11 198this account924median for 10K–100K
Reach, %39.62%this account3.62%median for 10K–100K
Engagement, %0.80%this account1.52%median for 10K–100K
| Metric | This account | Median for 10K–100K | Ratio |
|---|---|---|---|
| Median views per post | 11 198 | 924 | 12.1× |
| Reach (views ÷ followers) | 39.62% | 3.62% | 10.9× |
| Engagement rate | 0.80% | 1.52% | 0.52× |
Others in this range → Compare with another account → How these benchmarks are built →
Growth & engagement
How the posts we collected actually performed: views and reaction rate post by post, what the audience did with them, and where the follower count goes.
Views per post
2.7K28 May
1.6K
2.4K
2.5K
2.8K
1.3K29 May
29.5K4 Sep
20.1K12 Sep
16.8K13 Sep
1.4K
3K
33.4K
5.6K
2.6K
Last 14 collected posts, oldest on the left. The scale is logarithmic: one post can outrun the rest a hundred times over.
Engagement rate per post
0.59%28 May
0.80%
0.51%
0.36%
1.21%
0.47%29 May
0.44%4 Sep
0.63%12 Sep
1.06%13 Sep
0.96%
0.50%
2.29%
0.14%
1.95%
Reactions — likes, reposts, replies and quotes — divided by views. Median for 10K–100K accounts is 1.52%.
What the audience does
Likes70.8%2 223 in total
Reposts8.9%278 in total
Replies3.4%106 in total
Bookmarks17.0%534 in total
Share of every reaction we collected for this account. Replies mean argument, reposts mean endorsement, bookmarks mean the post was worth keeping.
The follower curve appears once this account has two daily snapshots — we take one a day, and this one is on its first.
Similar accounts
LEWINSKI ✓
@LEWINSKl
56.5K followers
· 22.6K posts
Twis Bren
@Bren_TM2
56.5K followers
· 1.7K posts
Young Africans SC English
@YoungAfricansEN
56.5K followers
· 7.7K posts
WZ | Wiener Zeitung
@WienerZeitung
56.5K followers
· 77.5K posts
world painting
@worldpainting2
56.5K followers
· 35.8K posts
Hens ✓
@Hens333
56.5K followers
· 6.8K posts
jacob
@jacobbmc2
56.5K followers
· 12.9K posts
X Lube Powder
@x_lube
56.5K followers
· 18.9K posts