MAROKO133 Update ai: Adobe Research Unlocking Long-Term Memory in Video World Models with

📌 MAROKO133 Update ai: Adobe Research Unlocking Long-Term Memory in Video World Mo

Video world models, which predict future frames conditioned on actions, hold immense promise for artificial intelligence, enabling agents to plan and reason in dynamic environments. Recent advancements, particularly with video diffusion models, have shown impressive capabilities in generating realistic future sequences. However, a significant bottleneck remains: maintaining long-term memory. Current models struggle to remember events and states from far in the past due to the high computational cost associated with processing extended sequences using traditional attention layers. This limits their ability to perform complex tasks requiring sustained understanding of a scene.

A new paper, “Long-Context State-Space Video World Models” by researchers from Stanford University, Princeton University, and Adobe Research, proposes an innovative solution to this challenge. They introduce a novel architecture that leverages State-Space Models (SSMs) to extend temporal memory without sacrificing computational efficiency.

The core problem lies in the quadratic computational complexity of attention mechanisms with respect to sequence length. As the video context grows, the resources required for attention layers explode, making long-term memory impractical for real-world applications. This means that after a certain number of frames, the model effectively “forgets” earlier events, hindering its performance on tasks that demand long-range coherence or reasoning over extended periods.

The authors’ key insight is to leverage the inherent strengths of State-Space Models (SSMs) for causal sequence modeling. Unlike previous attempts that retrofitted SSMs for non-causal vision tasks, this work fully exploits their advantages in processing sequences efficiently.

The proposed Long-Context State-Space Video World Model (LSSVWM) incorporates several crucial design choices:

1. Block-wise SSM Scanning Scheme: This is central to their design. Instead of processing the entire video sequence with a single SSM scan, they employ a block-wise scheme. This strategically trades off some spatial consistency (within a block) for significantly extended temporal memory. By breaking down the long sequence into manageable blocks, they can maintain a compressed “state” that carries information across blocks, effectively extending the model’s memory horizon.
2. Dense Local Attention: To compensate for the potential loss of spatial coherence introduced by the block-wise SSM scanning, the model incorporates dense local attention. This ensures that consecutive frames within and across blocks maintain strong relationships, preserving the fine-grained details and consistency necessary for realistic video generation. This dual approach of global (SSM) and local (attention) processing allows them to achieve both long-term memory and local fidelity.

The paper also introduces two key training strategies to further improve long-context performance:

• Diffusion Forcing: This technique encourages the model to generate frames conditioned on a prefix of the input, effectively forcing it to learn to maintain consistency over longer durations. By sometimes not sampling a prefix and keeping all tokens noised, the training becomes equivalent to diffusion forcing, which is highlighted as a special case of long-context training where the prefix length is zero. This pushes the model to generate coherent sequences even from minimal initial context.
• Frame Local Attention: For faster training and sampling, the authors implemented a “frame local attention” mechanism. This utilizes FlexAttention to achieve significant speedups compared to a fully causal mask. By grouping frames into chunks (e.g., chunks of 5 with a frame window size of 10), frames within a chunk maintain bidirectionality while also attending to frames in the previous chunk. This allows for an effective receptive field while optimizing computational load.

The researchers evaluated their LSSVWM on challenging datasets, including Memory Maze and Minecraft, which are specifically designed to test long-term memory capabilities through spatial retrieval and reasoning tasks.

The experiments demonstrate that their approach substantially surpasses baselines in preserving long-range memory. Qualitative results, as shown in supplementary figures (e.g., S1, S2, S3), illustrate that LSSVWM can generate more coherent and accurate sequences over extended periods compared to models relying solely on causal attention or even Mamba2 without frame local attention. For instance, on reasoning tasks for the maze dataset, their model maintains better consistency and accuracy over long horizons. Similarly, for retrieval tasks, LSSVWM shows improved ability to recall and utilize information from distant past frames. Crucially, these improvements are achieved while maintaining practical inference speeds, making the models suitable for interactive applications.

The Paper Long-Context State-Space Video World Models is on arXiv

The post Adobe Research Unlocking Long-Term Memory in Video World Models with State-Space Models first appeared on Synced.

🔗 Sumber: syncedreview.com

📌 MAROKO133 Hot ai: Scientists Spot Signs of Derelict Soviet Moon Lander on Lunar

In 1966, three years before the first humans walked on the Moon, the Soviet Union landed a small, spherical probe, dubbed Luna 9, on the lunar surface.

It was a historic moment, with the spacecraft becoming the first to achieve a soft landing and return the first photo from the surface of another celestial body: a high-contrast, black-and-white image of a rugged, rocky landscape.

Its actual whereabouts on the Moon, though, soon become a major point of contention. As the New York Times reports, two groups of scientists have now come forward saying they’ve found traces of the lost Soviet lander. However, the two groups don’t agree on where Luna 9 is hidden.

What makes it particularly difficult to spot is its diminutive size. Its spherical core stage measures a mere two feet across, making it roughly the size of a beach ball.

As Mark Robinson, principal investigator of the camera attached to NASA’s Lunar Reconnaissance Orbiter, told the NYT that Luna 9 is simply too small for the satellite to spot and confirm.

“You can stare at an image, and maybe that’s it, but you can’t really know for sure,” he said.

In a years-long effort to track down the historic relic, science communicator Vitaly Egorov turned to crowdsourcing to scan a 62-mile-wide region on the Moon for any unusual signs.

Egorov claims to have found Luna 9’s final resting place, after studying the horizon features in the grainy, black-and-white image the probe sent back to Earth 60 years ago.

He told the NYT that he’s “fairly confident” but admitted that he doesn’t “exclude an error of several meters.”

Fortunately, India’s space agency has agreed to use its higher-resolution satellite, Chandrayaan-2, to have a closer look in March.

Not everybody agrees with Egorov’s conclusion. Scientists at the University College London determined a different landing site, as detailed in a paper published in the journal npj Space Exploration last month.

The team came up with a machine-learning algorithm, inventively called “You-Only-Look-Once–Extraterrestrial Artefact” (YOLO-ETA), to train an AI on existing NASA findings of past landing sites on the Moon.

One spot, a bright pixel near two darker spots that could be Luna 9’s protective shells that it used to soften its landing, could be its final resting place, the University College London team claims.

For now, scientists await word from India’s Chandrayaan-2 team, which could soon shed more light on the matter — a treasure hunt over half a century in the making.

To some, it’s only a matter of time until we confirm where Luna 9 — and its twin, Luna 13 — have been resting for decades.

“It’s just a matter of placing bigger and better cameras into orbit around the moon,” Russian spaceflight expert and space journalist Anatoly Zak told NYT. “In our lifetimes, we probably will see those sites.”

More on spotting objects on the Moon: NASA Spots Object Speeding Around the Moon

The post Scientists Spot Signs of Derelict Soviet Moon Lander on Lunar Surface appeared first on Futurism.

🔗 Sumber: futurism.com