LingBot-Touch learns shared tactile tokens across sensor types
Authors say the tokens plug into diffusion policies and VLA models, including pi0.5 and LingBot-VLA.
Stories per day · 30 days
4 RSS
Relatedforecasting4time-series4flow-matching2pathfm2api1astrophysics1baidu1benchmarks1bert1cross-embodiment1dexterous-manipulation1ecg1embodied-ai1ernie1fev-bench1fine-tuning1fps-bench1gift-eval1
39 stories, newest first · one card per story
Stories tagged #foundation-model
Baidu releases ERNIE 5.5 Flash Base with 280 billion parameters
Open foundation model with 280 billion parameters and 14 billion active per input, pre-trained on 33.5 trillion tokens.
WeLM-Image unified image generation model integrates multimodal understanding
WeLM-Image combines text-to-image generation, instruction-guided editing and multi-reference synthesis on a single GPU with MoT backbone and MoE modules, ranking first on PRISM-Bench and Qwen-Image-Bench.
PrimeSim world model simulator from large-scale cross-embodiment demonstrations
PrimeSim is an action-conditioned multiview world model simulator that runs at 10.1 FPS and transfers to held-out platforms after 100 to 200 gradient steps.
Recursa recursively self-improving foundation model system
Recursa adapts to new domains by learning from its own experience and raises all-pass rates on held-out tasks from 30.0% to 55.0% with DeepSeek-V4.1-Flash.
Timestral releases four open time-series models, topping GIFT-Eval
Authors say the 209M and 123M variants rank first and second on every metric of GIFT-Eval, TIME, and fev-bench.
QECGBERT tokenizes ECG waveforms for BERT-style pretraining
It is pretrained on more than 800,000 unlabeled MIMIC-IV ECGs and fine-tuned on three labeled datasets.
OmniSpectra encodes spectra by wavelength instead of token rank
The 12.2M-parameter model, pretrained on eight surveys, beats seven baselines and larger models.
Mistral releases Timestral time-series foundation models
Mistral says its 209M model ranks first on GIFT-Eval, TIME and fev-bench while using about two-thirds the parameters of the prior leader.
PathFM: Fast Generative Forecasting of Multivariate Trajectories
PathFM generates full forecast trajectories with fast flow-matching sampling and leads a new path-score benchmark.
PathFM: Fast Generative Forecasting of Multivariate Trajectories
PathFM uses flow matching to forecast full multivariate trajectories and beats baselines on a new path-score benchmark.
Physical Intelligence API Beta: Frontier Models for Every Robot
Physical Intelligence opened a beta API to fine-tune and host its latest robot foundation models.
10
λ_0 humanoid model uses 500 hours of human loco-manipulation data
Zero-shot transfer to Unitree G1 reached 90.0% on SIMPLE and 62.5% on real loco-manipulation tasks.
Efficient Provably Private Classification with a Tabular Foundation Model
PrivTab is a differentially private tabular foundation model that classifies sensitive data in one forward pass.
Mirror Particle is building a ‘world model’ of human behavior
Mirror Particle is building a foundation world model to predict shifting human consumer behavior for brands.
5
NASA and IBM's open source lunar model turns 17 years of orbiter data into a foundation for lunar science
NASA and IBM released an open-source lunar foundation model trained on 17 years of orbiter data for ice and crater tasks.
Magic-W0: A Structured World-Action Foundation Model for Physical Intelligence
Magic-W0 jointly models structured 3D world transitions and robot actions, leading compared world-action models.
Higher-Order Molecular Grammars for Generative and Foundation Models in Chemistry
Higher-order grammar sequences let molecular models encode ring topology with 100% validity and stronger MoleculeNet transfer.
OpenTSLM TeeMoE: A Unified Time-Series Language Model for Forecasting, Contextual Prediction, and Reasoning
OpenTSLM TeeMoE unifies forecasting, contextual prediction, and temporal reasoning with a LoRA mixture of experts.
GroundingPI: A Grounding Foundation Model towards Physical Intelligence with Visual Primitives
GroundingPI, a 4B grounding model, sets a new state of the art across 34 benchmarks and improves robotics and driving.
GroundAnything: Reconciling Parallel Decoding with Precise Visual Grounding at Flash Speed
GroundAnything, a 4B model, pairs parallel diffusion decoding with precise visual grounding and large inference speedups.
TabFM: A Zero-Shot Foundation Model for Tabular Data
TabFM, a 400M tabular foundation model, delivers calibrated zero-shot predictions that beat tuned AutoML.
ALICE: In-context, Zero-shot, Mutual Information Estimation
ALICE estimates mutual information zero-shot from samples without retraining on each distribution.
Pretrain Once, Route Anywhere: Towards a Foundation Model for LLM Routing
RouteFM pretrains a frozen LLM router that generalizes to new models using behavioral context alone.