PPG-BENCH / EVALUATION
PDF PPG-Bench: A Unified Benchmark for PPG Foundation Models
A common protocol for evaluating representations from PPG.
Intelligence,beyond tokens.
Mannas is a frontier AI lab building models for language and the human brain.
Explore the lab01 / THE DIRECTION
Architected
for understanding.
Language research at Mannas moves beyond standard next-token prediction. In Sphota, a representation field connects context with expression, allowing an evolving understanding to shape what is communicated.
UNITY · INSIGHT · MANIFESTATION
Bhartṛhari’s conception of linguistic wholeness offers one influence: meaning held together, apprehended in context, and expressed. This is a computational inspiration, without equating a learned representation with his sphoṭa.
02 / MODELS
Built at Mannas.
03 / SELECTED RESEARCH
Models & resourcesLAST UPDATED · NEWEST FIRST
AXON / ATTENTION
Weights Adaptive Anisotropic Attention for Axis-Structured Signals
Learning how temporal and spatial attention work together.
PRATIBHA / CONREC
Weights Constrained Reconstruction for EEG Foundation Models
Shaping what a representation preserves through the reconstruction objective.
PRISM / TRANSFER
PDF PRISM: Exploring Heterogeneous Pretrained EEG Foundation Model Transfer to Clinical Differential Diagnosis
Testing how pretraining diversity changes clinical transfer.