CMU Machine Learning

Visual transition reasoning, not just diverse data, unlocks significant multimodal LLM improvements, with just 2,000 examples boosting 22 benchmarks.

Autonomous agents can recursively self-improve LLM-generated games, achieving superhuman performance and drastically cutting generation costs by internalizing development experience.

Multi-action weakly-coupled MDPs can shatter the theoretical $O(1/\sqrt{N})$ performance barrier to reach an $O(1/N)$ optimality gap simply by ditching heuristic priority ordering in favor of locally linear mean-field dynamics.

It is shown that the workspace token can be used as a drop-in replacement for observations during deployment, enabling policies to solve memory-intensive tasks without the need for VLM reasoning in-the-loop.

Achieving higher novel-view quality in 3D scene generation without requiring dense volumetric representations or ground-truth supervision is a game changer for practical applications.

Agents fall short in machine learning development, often locked in narrow loops while humans adapt and innovate across tasks.

Group Decentralized RHCR achieves high throughput for multi-agent pathfinding while slashing computational costs, making it viable for larger agent counts.

Prediction objectives shape the physical knowledge embedded in latent representations, revealing that some parameters are only captured under specific forecasting conditions.

Achieving over 99% fulfillment of observation requests in decentralized satellite scheduling reveals a transformative approach to large-scale DCOPs that outshines traditional methods.

Real-time manipulation policies can now react to sensory input every 40ms, achieving up to a 30% boost in success rates in dynamic environments.

PatternGSL achieves accurate garment reconstruction from images without templates, revolutionizing how we think about garment design and simulation.

Generalist agents need to remember distinct information to navigate conflicting optimal actions across environments, challenging the notion that current state observations are sufficient.

MCPS reveals that analyzing football passes through a Monte Carlo framework can significantly improve the understanding of pass effectiveness and decision-making in real-time scenarios.

iOSWorld reveals that even state-of-the-art models falter in multi-app reasoning, achieving only 37% accuracy, underscoring the complexity of personal context in AI interactions.

Test-time performance can be significantly improved without a corresponding drop in validation loss, raising critical questions about model evaluation in deep learning.

ALMANAC reveals that agents can significantly improve their collaborative competence by learning from detailed human mental model annotations.

Multi-agent coordination can boost task completion efficiency by up to 25.5%, transforming how we approach complex computer use challenges.

AHEAD transforms static VLA models into dynamic manipulators, achieving up to 97% success in scenarios where traditional methods fail spectacularly.

Steering imaginations in video world models can reveal critical failure points in robotic actions that traditional methods might overlook.

Achieve SOTA reasoning performance with a 30B parameter model using up to 95% fewer tokens by explicitly controlling when and how deeply an LLM plans.

Control over physical properties like friction and restitution in generated videos is now possible, paving the way for more realistic and controllable video synthesis.

Existing robotic methods falter in tackling fundamental physical reasoning challenges, as evidenced by KinDER's rigorous benchmark evaluation.

Extracting temporal geometry from generative models can boost reinforcement learning performance by over 2x without changing the optimal policy.

Humanoid robots can now perform complex, contact-rich manipulation tasks with significantly improved dexterity and success by "dreaming" about the feel of their actions.

Forget scraping the internet for physics Q&A: this work shows you can train LLMs to ace the International Physics Olympiad using only data generated from physics simulators.

By fusing IMU and insole pressure data within a physics simulation, GRIP achieves more physically plausible human motion capture than IMU-only methods.

Injecting muscle synergy priors into reinforcement learning drastically improves the realism of simulated human locomotion, even with limited real-world data.

Unsupervised discovery of object keypoints and dynamics directly from video unlocks state-of-the-art world models applicable to decision-making.

Robots can now learn long-horizon tasks far more effectively by distilling complex histories into a few key visual moments, outperforming standard imitation learning by 70% on real-world tasks.