Loading video...

Video Failed to Load

There was a problem loading this video. This could be due to a temporary network issue or the video might be unavailable.

Ego2Web A Web Agent Benchmark Grounded in Egocentric Videos paper:

AK

511,941 subscribers

17,070 views • 4 months ago •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

0 Comments

No comments available

Comments from the original post will appear here

Related Videos

Our paper “Learning Situated Awareness in the Real World” has been accepted to #ICML2026 as a Spotlight (top 2.2%)! Congrats to Chuhan Li ✈️ICML and the team! We introduce SAW-Bench, a real-world egocentric benchmark for evaluating situated (first-person) spatial reasoning in multimodal models, uncovering a large gap between humans and current systems. All videos recorded via Meta Ray-Ban 2 glasses!

Our paper “Learning Situated Awareness in the Real World” has been accepted to #ICML2026 as a Spotlight (top 2.2%)! Congrats to Chuhan Li ✈️ICML and the team! We introduce SAW-Bench, a real-world egocentric benchmark for evaluating situated (first-person) spatial reasoning in multimodal models, uncovering a large gap between humans and current systems. All videos recorded via Meta Ray-Ban 2 glasses!

Xin Eric Wang

11,529 views • 3 months ago

EgoLifter Open-world 3D Segmentation for Egocentric Perception In this paper we present EgoLifter, a novel system that can automatically segment scenes captured from egocentric sensors into a complete decomposition of individual 3D objects. The system is specifically

EgoLifter Open-world 3D Segmentation for Egocentric Perception In this paper we present EgoLifter, a novel system that can automatically segment scenes captured from egocentric sensors into a complete decomposition of individual 3D objects. The system is specifically

AK

41,117 views • 2 years ago

1/ [NeurIPS D&B] Introducing HourVideo: A benchmark for hour-long video-language understanding!🚀 500 egocentric videos, 18 total tasks & ~13k questions! Performance: GPT-4➡️25.7% Gemini 1.5 Pro➡️37.3% Humans➡️85.0% We highlight a significant gap in multimodal capabilities🧵👇

1/ [NeurIPS D&B] Introducing HourVideo: A benchmark for hour-long video-language understanding!🚀 500 egocentric videos, 18 total tasks & ~13k questions! Performance: GPT-4➡️25.7% Gemini 1.5 Pro➡️37.3% Humans➡️85.0% We highlight a significant gap in multimodal capabilities🧵👇

Keshigeyan Chandrasegaran

111,369 views • 1 year ago

🛑📢 HD-EPIC: A Highly-Detailed Egocentric Video Dataset New collected videos 263 annotations/min: recipe, nutrition, actions, sounds, 3D object movement &fixture associations, masks. 26K VQA benchmark to challenge current VLMs 1/N

🛑📢 HD-EPIC: A Highly-Detailed Egocentric Video Dataset New collected videos 263 annotations/min: recipe, nutrition, actions, sounds, 3D object movement &fixture associations, masks. 26K VQA benchmark to challenge current VLMs 1/N

Dima Damen

22,320 views • 1 year ago

How capable are web agents at solving knowledge work tasks? 🤔 Are LLMs up to the challenge? 🤖 Introducing WorkArena: a benchmark where agents meet the world 𝘸𝘪𝘭𝘥 web of enterprise software 🌐🖥️ Paper: Website: 🧵 1/7

How capable are web agents at solving knowledge work tasks? 🤔 Are LLMs up to the challenge? 🤖 Introducing WorkArena: a benchmark where agents meet the world 𝘸𝘪𝘭𝘥 web of enterprise software 🌐🖥️ Paper: Website: 🧵 1/7

Alexandre Lacoste

24,536 views • 2 years ago

we are kicking off Paper video tutorials in a big way! upcoming videos will cover our pen tool and tokens launches and new agent model explorations what else would you like to see?

we are kicking off Paper video tutorials in a big way! upcoming videos will cover our pen tool and tokens launches and new agent model explorations what else would you like to see?

Paper

28,663 views • 2 months ago

wow... GPT 5.6 Sol Passed the Replit Benchmark. In one prompt 5.6 on Codex created a full Replit clone with DB, Sandboxes, and an agent that builds and edits full web apps. Only Fable and GPT 5.6 have passed this benchmark. 5.6 is much better at designing iOS apps than 5.5 as well.

wow... GPT 5.6 Sol Passed the Replit Benchmark. In one prompt 5.6 on Codex created a full Replit clone with DB, Sandboxes, and an agent that builds and edits full web apps. Only Fable and GPT 5.6 have passed this benchmark. 5.6 is much better at designing iOS apps than 5.5 as well.

Riley Brown

123,242 views • 27 days ago

1️⃣ Ego-Exo4D A new foundational dataset + benchmark suite to support research on video learning & multimodal perception, co-developed with 14 university partners. Details ➡️ Core to the work is videos of skilled human activities, simultaneously capturing both first-person "egocentric" + multiple “exocentric” views.

1️⃣ Ego-Exo4D A new foundational dataset + benchmark suite to support research on video learning & multimodal perception, co-developed with 14 university partners. Details ➡️ Core to the work is videos of skilled human activities, simultaneously capturing both first-person "egocentric" + multiple “exocentric” views.

AI at Meta

133,666 views • 2 years ago

Today we’re releasing Ramp SWE-Bench: a private, production-grounded coding benchmark created from real engineering problems we've faced at Ramp.

Today we’re releasing Ramp SWE-Bench: a private, production-grounded coding benchmark created from real engineering problems we've faced at Ramp.

Ramp Labs

186,116 views • 1 month ago

Paper is now a canvas for Cursor, Claude Code, Codex. Any agent can read and write to your design.

Paper is now a canvas for Cursor, Claude Code, Codex. Any agent can read and write to your design.

Paper

12,452,561 views • 4 months ago

From egocentric glove data to a MANO hand model, replayed in Rerun.

From egocentric glove data to a MANO hand model, replayed in Rerun.

Yu Xiang

39,229 views • 6 months ago

Meta PARTNR is a benchmark for planning and reasoning in embodied multi-agent tasks. This large-scale human and robot collaboration benchmark was core to our recent demos and also informs our work as scientists and engineers pushing this field of study forward.

Meta PARTNR is a benchmark for planning and reasoning in embodied multi-agent tasks. This large-scale human and robot collaboration benchmark was core to our recent demos and also informs our work as scientists and engineers pushing this field of study forward.

AI at Meta

33,607 views • 1 year ago

Anything you can do in Notion, your agent can do… on mobile. Before, Agent on mobile ran in a web view. Now, it’s native, faster, so quick ideas don’t have to wait.

Anything you can do in Notion, your agent can do… on mobile. Before, Agent on mobile ran in a web view. Now, it’s native, faster, so quick ideas don’t have to wait.

Notion

60,969 views • 7 months ago

Sprint planner agent: don’t create/move your tickets manually, let your agent take care of them Designed in Paper, animation with Motion inside Zed

Sprint planner agent: don’t create/move your tickets manually, let your agent take care of them Designed in Paper, animation with Motion inside Zed

Jeet

31,129 views • 2 months ago

pi-web-access is an extension that provides web search (and more) for Pi coding agent. The latest update introduces a new UX pattern: you can hit `Ctrl+S` to intercept agent searches mid-flight to inspect and curate exactly what makes it into the agent conversation. pi install npm:pi-web-access

pi-web-access is an extension that provides web search (and more) for Pi coding agent. The latest update introduces a new UX pattern: you can hit `Ctrl+S` to intercept agent searches mid-flight to inspect and curate exactly what makes it into the agent conversation. pi install npm:pi-web-access

Nico Bailon

29,352 views • 5 months ago

New drop in the Agent Spotlight series. Want to deploy a web app just by sending a message? The n8n.io-Vercel agent on ASI:One lets you go from repo → live URL in seconds. ⚡ No setup. Just ask. Try it on ASI:One: Explore the agent:

New drop in the Agent Spotlight series. Want to deploy a web app just by sending a message? The n8n.io-Vercel agent on ASI:One lets you go from repo → live URL in seconds. ⚡ No setup. Just ask. Try it on ASI:One: Explore the agent:

Fetch.ai

14,074 views • 1 year ago

Existing IR/RAG benchmarks are unrealistic: they’re often derived from easily retrievable topics, rather than grounded in solving real user problems. 🧵Introducing 𝐅𝐫𝐞𝐬𝐡𝐒𝐭𝐚𝐜𝐤, a challenging RAG benchmark on niche, recent topics. Work done during intern Databricks 🧱

Existing IR/RAG benchmarks are unrealistic: they’re often derived from easily retrievable topics, rather than grounded in solving real user problems. 🧵Introducing 𝐅𝐫𝐞𝐬𝐡𝐒𝐭𝐚𝐜𝐤, a challenging RAG benchmark on niche, recent topics. Work done during intern Databricks 🧱

Nandan Thakur

40,063 views • 1 year ago

Announcing Agent S3 — our most advanced computer-use agent yet — achieving 69.9% on OSWorld, just shy of the 72% human benchmark. Today also marks one year since Simular launched the first generation of Agent S. From 20% to 69.9% in just twelve months. What a year. Let that sink in. 🧠💻 🧵👇

Announcing Agent S3 — our most advanced computer-use agent yet — achieving 69.9% on OSWorld, just shy of the 72% human benchmark. Today also marks one year since Simular launched the first generation of Agent S. From 20% to 69.9% in just twelve months. What a year. Let that sink in. 🧠💻 🧵👇

Ang Li

444,496 views • 10 months ago

The way this Paper Mario dialogue instantly activated me like a sleeper agent is INSANE

Sensitive content

The way this Paper Mario dialogue instantly activated me like a sleeper agent is INSANE

Sam

17,261 views • 2 years ago

How well do today’s frontier models handle long-horizon, multi-step web agent tasks, such as identifying the top 25 U.S. CS PhD programs with ML/AI faculty likely accepting students and compiling the results into a structured sheet? Check out our new work on Odysseys: Benchmarking Web Agents on Realistic Long Horizon Tasks Paper: Leaderboard: We introduce Odysseys, a benchmark of 200 long-horizon tasks derived from real browsing sessions and evaluated on the live Internet. We show that binary pass/fail is inadequate in this setting and propose rubric-based evaluation, which better aligns with human judgment and provides more informative signals. Across leading models, the best achieves only 44.5% success, leaving substantial headroom. We further introduce a Trajectory Efficiency metric (rubric score per step) and find efficiency remains extremely low (1.15%), highlighting a key bottleneck. Odysseys provides a realistic benchmark for measuring progress toward web agents capable of sustained, efficient, real-world operation. See a more detailed thread by Jing Yu Koh.

How well do today’s frontier models handle long-horizon, multi-step web agent tasks, such as identifying the top 25 U.S. CS PhD programs with ML/AI faculty likely accepting students and compiling the results into a structured sheet? Check out our new work on Odysseys: Benchmarking Web Agents on Realistic Long Horizon Tasks Paper: Leaderboard: We introduce Odysseys, a benchmark of 200 long-horizon tasks derived from real browsing sessions and evaluated on the live Internet. We show that binary pass/fail is inadequate in this setting and propose rubric-based evaluation, which better aligns with human judgment and provides more informative signals. Across leading models, the best achieves only 44.5% success, leaving substantial headroom. We further introduce a Trajectory Efficiency metric (rubric score per step) and find efficiency remains extremely low (1.15%), highlighting a key bottleneck. Odysseys provides a realistic benchmark for measuring progress toward web agents capable of sustained, efficient, real-world operation. See a more detailed thread by Jing Yu Koh.

Russ Salakhutdinov

22,518 views • 3 months ago