Pinned Tweet
🌟 Excited to share that our paper, RoboReason, has been accepted to EMNLP 2026!
🤖 We study how well Vision-Language Models (VLMs) perform on spatiotemporal ordering tasks that use real-world robotics images.
Our key findings: 🔹 Large VLMs often underperform previously valuated on spatiotemporal ordering tasks using real-world robotics images. 🔹 Overlaying specialized computer vision model outputs and increasing reasoning effort generally help improve large VLM performance. 🔹 Finetuning smaller models using GRPO and SFT on examples generated by specialized computer vision models often helps them achieve comparable results to much larger VLMs.
📍 Presenting with @sagnikanupam at @emnlpmeeting in Budapest 🇭🇺 tomorrow!
🗓️ Poster Session C | 9:00–10:30 AM
#EMNLP2026 #VLMs #LLMs #Robotics #AI
🤖 We study how well Vision-Language Models (VLMs) perform on spatiotemporal ordering tasks that use real-world robotics images.
Our key findings: 🔹 Large VLMs often underperform previously valuated on spatiotemporal ordering tasks using real-world robotics images. 🔹 Overlaying specialized computer vision model outputs and increasing reasoning effort generally help improve large VLM performance. 🔹 Finetuning smaller models using GRPO and SFT on examples generated by specialized computer vision models often helps them achieve comparable results to much larger VLMs.
📍 Presenting with @sagnikanupam at @emnlpmeeting in Budapest 🇭🇺 tomorrow!
🗓️ Poster Session C | 9:00–10:30 AM
#EMNLP2026 #VLMs #LLMs #Robotics #AI
13
13
36
1,530
Utkarsh ✈️ EMNLP 2026 retweeted
🌟 Excited to share that our paper, RoboReason, has been accepted to EMNLP 2026!
🤖 We study how well Vision-Language Models (VLMs) perform on spatiotemporal ordering tasks that use real-world robotics images.
Our key findings: 🔹 Large VLMs often underperform previously valuated on spatiotemporal ordering tasks using real-world robotics images. 🔹 Overlaying specialized computer vision model outputs and increasing reasoning effort generally help improve large VLM performance. 🔹 Finetuning smaller models using GRPO and SFT on examples generated by specialized computer vision models often helps them achieve comparable results to much larger VLMs.
📍 Presenting with @sagnikanupam at @emnlpmeeting in Budapest 🇭🇺 tomorrow!
🗓️ Poster Session C | 9:00–10:30 AM
#EMNLP2026 #VLMs #LLMs #Robotics #AI
🤖 We study how well Vision-Language Models (VLMs) perform on spatiotemporal ordering tasks that use real-world robotics images.
Our key findings: 🔹 Large VLMs often underperform previously valuated on spatiotemporal ordering tasks using real-world robotics images. 🔹 Overlaying specialized computer vision model outputs and increasing reasoning effort generally help improve large VLM performance. 🔹 Finetuning smaller models using GRPO and SFT on examples generated by specialized computer vision models often helps them achieve comparable results to much larger VLMs.
📍 Presenting with @sagnikanupam at @emnlpmeeting in Budapest 🇭🇺 tomorrow!
🗓️ Poster Session C | 9:00–10:30 AM
#EMNLP2026 #VLMs #LLMs #Robotics #AI
13
13
36
1,530
Utkarsh ✈️ EMNLP 2026 retweeted
🌟 Excited to share that our paper, RoboReason, has been accepted to EMNLP 2026!
🤖 We study how well Vision-Language Models (VLMs) perform on spatiotemporal ordering tasks that use real-world robotics images.
Our key findings:
🔹 Large VLMs often underperform previously valuated on spatiotemporal ordering tasks using real-world robotics images.
🔹 Overlaying specialized computer vision model outputs and increasing reasoning effort generally help improve large VLM performance.
🔹 Finetuning smaller models using GRPO and SFT on examples generated by specialized computer vision models often helps them achieve comparable results to much larger VLMs.
📍 Presenting in Budapest with @utk7arsh at @emnlpmeeting, tomorrow!
🗓️ Poster Session C | 9:00–10:30 AM
#EMNLP2026 #VLMs #LLMs #Robotics #AI
🤖 We study how well Vision-Language Models (VLMs) perform on spatiotemporal ordering tasks that use real-world robotics images.
Our key findings:
🔹 Large VLMs often underperform previously valuated on spatiotemporal ordering tasks using real-world robotics images.
🔹 Overlaying specialized computer vision model outputs and increasing reasoning effort generally help improve large VLM performance.
🔹 Finetuning smaller models using GRPO and SFT on examples generated by specialized computer vision models often helps them achieve comparable results to much larger VLMs.
📍 Presenting in Budapest with @utk7arsh at @emnlpmeeting, tomorrow!
🗓️ Poster Session C | 9:00–10:30 AM
#EMNLP2026 #VLMs #LLMs #Robotics #AI
2
3
12
455
📄 Read the paper: sagnikanupam.com/papers/roborea…
Huge thanks to my amazing collaborators: @sagnikanupam, Aneesh Edara, Botong Zhang, @tyao923, @thoma_gu, @dineshjayaraman, and @obastani!
Huge thanks to my amazing collaborators: @sagnikanupam, Aneesh Edara, Botong Zhang, @tyao923, @thoma_gu, @dineshjayaraman, and @obastani!
0
0
1
96
Dropping paper soon. Can’t wait to show what we have been working on and presenting at #EMNLP Budapest 🇭🇺 this week.
Meanwhile great seeing a bunch of papers and meeting new people at my first in person conference.
Meanwhile great seeing a bunch of papers and meeting new people at my first in person conference.
2
0
27
744
Utkarsh ✈️ EMNLP 2026 retweeted
Your phone is full of apps built to farm your ATTENTION
i want the other kind. so i started World's Smallest Studio, and today it became a real org.
the kind of software it stands for:
- goes where you go, from your phone to your desktop
- full quality, because privacy is no pp
- yours for life: your data, your identity, your workflow
- no ads, no lock in, no noise
step one is on the desk: smallfil, one file library synced across android and desktop, fixing passwords, backup codes next. then a family of tools sharing one identity
but this shouldn't be one studio. it should be a move
it you want to ship software like this, small, calm and excellent reply or DM me, Let's build it together
i want the other kind. so i started World's Smallest Studio, and today it became a real org.
the kind of software it stands for:
- goes where you go, from your phone to your desktop
- full quality, because privacy is no pp
- yours for life: your data, your identity, your workflow
- no ads, no lock in, no noise
step one is on the desk: smallfil, one file library synced across android and desktop, fixing passwords, backup codes next. then a family of tools sharing one identity
but this shouldn't be one studio. it should be a move
it you want to ship software like this, small, calm and excellent reply or DM me, Let's build it together
11
2
17
472
Vision-force-action model 🤔🫡
I have been reading about eka which targets the core robotics tradeoff by using force sensing, reinforcement learning in simulation, and physical execution to deliver superhuman performance, achieving 1,500 picks per hour at over 95% reliability on diverse objects up to 1kg.
Curious to see who their industrial customers are and what they have to see about the performance vs generality discussion.
Another big player turning up in the market 🔥
I have been reading about eka which targets the core robotics tradeoff by using force sensing, reinforcement learning in simulation, and physical execution to deliver superhuman performance, achieving 1,500 picks per hour at over 95% reliability on diverse objects up to 1kg.
Curious to see who their industrial customers are and what they have to see about the performance vs generality discussion.
Another big player turning up in the market 🔥
@EkaRobotics has raised a $150M Series A.
Generality v/s Performance is the fundamental tension at the frontier of robotics. Generality without performance is just a prototype. Performance without generality is just a single-purpose tool.
Customers shouldn’t have to choose. Eka wered by our Vision-Force-Action (VFA) model, we deliver the performance and versatility customers need to put robots to work every day.
Eka is built on three convictions:
⚡️Force is the language of manipulation.
🚀 Reinforcement learning is how dexterous manipulation scales.
🎂 System 1—fast, reliable physical execution—is the bottleneck to solve.
With this funding, we will scale our model and accelerate customer deployments.
Grateful to be building alongside a dream team and investors who share our ambition.
0
0
22
891
I have seen @DetectIfAI ‘s journey and progress since December and with its new drop, all of us can get access to what Tarini and her team has been building.
Very strong numbers in latency and benchmark pairs well with concrete production impact from other major vertical players and this technology is now in developers’ hands.
I’m most excited about checking out performance against cloned voice
Very strong numbers in latency and benchmark pairs well with concrete production impact from other major vertical players and this technology is now in developers’ hands.
I’m most excited about checking out performance against cloned voice
today we're opening @DetectIfAI to every developer.
real-time voice deepfake detection, 24 ms per check.
#1 in latency on Podonos' independent benchmark, ahead of companies that raised hundreds of millions.
97.3% of cloned voices caught.
API + MCP. link in the comments.
3
0
13
520
First time using @PhotonHQ. Might be cooking something good. Stay tuned
0
0
9
406
As I was preparing to retire from hackathons, Claude just had to get so much better that I physically can’t atp 🥹🥀
3
1
19
609
Here comes another one of good old “gpt moment for robotics.”
Announcing the beta release of the Pi API. We’ve partnered with companies across diverse industries to adapt our models to their robots and applications, from harvesting strawberries to assembling servers. Now we’re opening access to more teams.
1
0
14
614
Utkarsh ✈️ EMNLP 2026 retweeted
i kept complaining to my grok bot about my browser... so it went off and built me one.
@ziwenxu_ and i took it from there. now Serein is live.
an AGENTIC HARNESS for your browser
>chat with claude code or codex right inside it
>let them control your tabs and do the work
>use the toolst's free. your browser is their workspace now.
links in the comments check it out👇
@ziwenxu_ and i took it from there. now Serein is live.
an AGENTIC HARNESS for your browser
>chat with claude code or codex right inside it
>let them control your tabs and do the work
>use the toolst's free. your browser is their workspace now.
links in the comments check it out👇
44
21
91
17,367
Utkarsh ✈️ EMNLP 2026 retweeted
Closing out SF Tech Week at Speed Hacks 🏁 Let's ship.
@speedrun @a16z @HackerSquad #SpeedHacks #SFTechWeek #SFHackWeek #speedrun #buildinpublic
@speedrun @a16z @HackerSquad #SpeedHacks #SFTechWeek #SFHackWeek #speedrun #buildinpublic
7
3
19
400