Bottleneck Tokens paired with a masked generative objective achieve state-of-the-art unified multimodal retrieval performance among 2B-scale models on the MMEB-V2 benchmark with 78 datasets.
Canonical reference
Perceiver: General perception with iterative attention
Canonical reference. 80% of citing Pith papers cite this work as background.
citation-role summary
citation-polarity summary
representative citing papers
LGS pretrained on 2.5M trajectories across 16 systems matches deterministic baselines at one step and halves 20-step error while using far less compute and adapting to held-out higher-resolution flows.
Self-supervised Perceiver-VAE pre-trained on 227,000 light curves from MMT-9 and fine-tuned on simulators achieves 85% accuracy and 0.92-0.95 ROC AUC in anomaly detection and motion mode prediction for space objects.
StereoPolicy fuses stereo image pairs via a Stereo Transformer on pretrained 2D encoders to boost robotic manipulation policies, showing gains over monocular, RGB-D, point cloud, and multi-view methods in simulations and real-robot tests.
MetaRL pre-trained on GBWM problems delivers near-optimal dynamic strategies in 0.01s achieving 97.8% of DP optimal utility and handles larger problems where DP fails.
Knowledge conflicts in hypernetwork LLM adaptation stem from constant adapter margins losing to frequency-dependent pretrained margins; selective layer boosting and conflict-aware triggering raise deep-conflict accuracy to 71-72.5% on Gemma-2B and Mistral-7B.
HubRouter is a sub-quadratic routing primitive using learned hubs that replaces attention layers in hybrid models while delivering competitive perplexity and large throughput gains.
Transformer models applied to simulated water-Cherenkov array data improve gamma-hadron separation and reconstruction of direction, core position, and energy compared to established techniques.
GAF creates 4D dynamic scene models by adding motion to 3D Gaussians, enabling better reconstruction and 7.3% higher success in robotic tasks.
citing papers explorer
-
Bottleneck Tokens for Unified Multimodal Retrieval
Bottleneck Tokens paired with a masked generative objective achieve state-of-the-art unified multimodal retrieval performance among 2B-scale models on the MMEB-V2 benchmark with 78 datasets.
-
Latent Generative Solvers for Generalizable Long-Term Physics Simulation
LGS pretrained on 2.5M trajectories across 16 systems matches deterministic baselines at one step and halves 20-step error while using far less compute and adapting to held-out higher-resolution flows.
-
A Self-Supervised Framework for Space Object Behaviour Characterisation
Self-supervised Perceiver-VAE pre-trained on 227,000 light curves from MMT-9 and fine-tuned on simulators achieves 85% accuracy and 0.92-0.95 ROC AUC in anomaly detection and motion mode prediction for space objects.
-
StereoPolicy: Improving Robotic Manipulation Policies via Stereo Perception
StereoPolicy fuses stereo image pairs via a Stereo Transformer on pretrained 2D encoders to boost robotic manipulation policies, showing gains over monocular, RGB-D, point cloud, and multi-view methods in simulations and real-robot tests.
-
A Meta Reinforcement Learning Approach to Goals-Based Wealth Management
MetaRL pre-trained on GBWM problems delivers near-optimal dynamic strategies in 0.01s achieving 97.8% of DP optimal utility and handles larger problems where DP fails.
-
The Override Gap: A Magnitude Account of Knowledge Conflict Failure in Hypernetwork-Based Instant LLM Adaptation
Knowledge conflicts in hypernetwork LLM adaptation stem from constant adapter margins losing to frequency-dependent pretrained margins; selective layer boosting and conflict-aware triggering raise deep-conflict accuracy to 71-72.5% on Gemma-2B and Mistral-7B.
-
HubRouter: A Pluggable Sub-Quadratic Routing Primitive for Hybrid Sequence Models
HubRouter is a sub-quadratic routing primitive using learned hubs that replaces attention layers in hybrid models while delivering competitive perplexity and large throughput gains.
-
Enhancing event reconstruction for $\gamma$-ray particle detector arrays using transformers
Transformer models applied to simulated water-Cherenkov array data improve gamma-hadron separation and reconstruction of direction, core position, and energy compared to established techniques.
-
GAF: Gaussian Action Field as a 4D Representation for Dynamic World Modeling in Robotic Manipulation
GAF creates 4D dynamic scene models by adding motion to 3D Gaussians, enabling better reconstruction and 7.3% higher success in robotic tasks.