Meta tags:

Headings (most frequently used words):

world, to, the, genex, explore, their, beliefs, generative, explorer, demo, with, challenge, embodied, ai, of, this, that, physically, update, about, can, exploration, and, imagined, observations, watch, our, introduction, planning, partial, observation, is, central, in, majority, prior, works, have, tackled, by, developing, agents, environment, state, however, humans, imagine, unseen, parts, through, mental, revise, such, updated, allow, them, make, more, informed, decisions, at, current, step, without, having, first, achieve, human, like, ability, we, introduce, video, generation, model, allows, an, agent, mentally, large, scale, 3d, urban, scenes, acquire, its, belief, architecture, navigation, consistency, for, interactive,

Text of the page (most frequently used words):
the (21), world (11), genex (9), and (9), video (9), you (7), can (7), for (5), agent (5), with (5), that (5), our (5), #explorer (5), your (4), support (4), embodied (4), see (4), their (4), panoramic (4), forward (4), image (4), generative (4), browser (3), does (3), not (3), agents (3), pedestrian (3), car (3), views (3), this (3), generation (3), introduce (3), consistency (3), exploration (3), diffuser (3), movement (3), through (3), architecture (3), explore (3), beliefs (3), interactive (2), demo (2), scenario (2), involving (2), making (2), street (2), but (2), make (2), are (2), them (2), tag (2), path (2), input (2), like (2), navigation (2), model (2), environment (2), ropf (2), challenge (2), physically (2), update (2), about (2), imagined (2), observations (2), introduction (2), html5, canvas, element, gives, new, level, intelligence, single, decision, picture, yourself, driving, down, when, siren, sounds, source, project, what, ahead, reveals, ambulance, just, around, corner, stop, crossing, space, pass, multi, interaction, suppose, waiting, red, light, approaching, everything, seems, fine, helps, recognize, likewise, vehicle, blocking, realizing, act, immediately, warn, preventing, collision, ensure, quality, navigational, cycle, navigates, randomly, sampled, closed, returning, origin, optimal, cases, start, end, identical, ensuring, modeling, trained, seamlessly, adapts, generating, along, from, use, lmms, gpt, controlling, egocentric, view, move, change, direction, generated, generator, angle, changes, handled, perform, unlimited, actions, design, variant, takes, height, width, outputs, sequence, images, predicted, aim, generate, realistic, 360, degree, sequences, simulate, any, leveraging, diffusion, models, designed, spherical, consistent, learning, navigataion, planning, partial, observation, central, majority, prior, works, have, tackled, developing, state, however, humans, imagine, unseen, parts, mental, revise, such, updated, allow, more, informed, decisions, current, step, without, having, first, achieve, human, ability, allows, mentally, large, scale, urban, scenes, acquire, its, belief, watch, taiming, tianmin, shu, alan, yuille, daniel, khashabi, jieneng, chen, evaluation,

Text of the page (random words):
generative world explorer genex video introduction architecture navigation evaluation embodied ai interactive generative world explorer taiming lu tianmin shu alan yuille daniel khashabi jieneng chen watch our demo your browser does not support the video tag introduction planning with partial observation is a central challenge in embodied ai a majority of prior works have tackled this challenge by developing agents that physically explore their environment to update their beliefs about the world state however humans can imagine unseen parts of the world through a mental exploration and revise their beliefs with imagined observations such updated beliefs can allow them to make more informed decisions at the current step without having to physically explore the world first to achieve this human like ability we introduce the generative world explorer genex a video generation model that allows an agent to mentally explore a large scale 3d world e g urban scenes and acquire imagined observations to update its belief about the world genex architecture to support exploration we design a variant of video generation model f θ ropf h w ropf t h w that takes an input image with height h and width w and outputs a sequence of t images as predicted video we aim to generate realistic 360 degree panoramic video sequences that simulate forward movement through any environment by leveraging image to video diffusion models f θ we introduce our diffuser architecture and designed spherical consistent learning for panoramic navigataion genex navigation our trained diffuser seamlessly adapts as a world explorer generating panoramic views along a forward path from an image input we use lmms like gpt 4o as agents controlling movement through an egocentric view the agent can move forward or change direction with forward movement generated by our video generator and angle changes handled by the panoramic image the agent can perform unlimited actions making our diffuser a generative world explorer exploration consistency to ensure generation quality we introduce navigational cycle consistency genex navigates a randomly sampled closed path returning to the origin in optimal cases the start and end views are identical ensuring consistency in world modeling genex for embodied ai your browser does not support the video tag genex gives a new level of intelligence to embodied ai agents for a single agent scenario involving decision making picture yourself driving down a street when a siren sounds but you can t see the source with genex you project what s ahead it reveals an ambulance just around the corner you stop crossing the street to make space for it to pass for a multi agent scenario involving interaction suppose you are waiting at a red light you see a pedestrian and an approaching car and everything seems fine but genex helps you recognize that the pedestrian can t see the car likewise the car can t see the pedestrian your vehicle is blocking their views realizing this you act immediately to warn them preventing a collision interactive demo your browser does not support the html5 canvas element