Deux Voir Deux Voir

Independent researchers · est. 2026

World models, AI Safety
& Applied AI

Deux Voir is a small, independent research team and identity investigating the topics of learned world-models; their latent representations, and LLM inferencing, fine-tuning.

Selected work

Under review Venue TBD

World Models as Neural Surrogates
for DFT-Scale Materials Property Simulation

Arjun Srivastava · Gaurav Karakoti

A two-stage neural surrogate for high-throughput materials discovery: a JEPA encoder learns physics-informed latent representations of crystal graph structure via masked embedding prediction, and a lightweight transformer readout predicts formation energy, bandgap, and stability — without any quantum-mechanical labels and without performing DFT calculations at inference time.

Read paper
Workshop submission TBD

Concurrency for Accelerating Multi-Modal
Tensor Inference on Resource-Constrained Architectures

Arjun Srivastava

DCTMS is a compile-time framework that analyses a transformer's computational graph before inference and derives a complete tensor lifetime map, physical buffer assignment, and concurrency schedule. On Apple M4 unified memory, it reduces 448 logical tensors to 122 physical buffers (72.77%) and establishes a provable memory envelope before inference begins, cutting latency variance by 55–69% at the 7B scale.

Read paper
Workshop submission ECCV 2026 · Wearable AI

World-Model Latents as Implicit Memory
for Egocentric Video Reasoning

Arjun Srivastava · Gaurav Karakoti

We investigate whether the learned latent representations of a visual world model can serve as a compact memory substrate for long-context egocentric video reasoning. A V-JEPA encoder produces persistent latent state; a small Qwen vision-language model reasons over that state directly — without external retrieval.

Read paper PDF

More work such as models, tools, and experiments will appear here as it becomes ready.

Peer Reviewing Contribution

We have peer reviewing for the following conferences and journals.

Arjun Srivastava

  • NeurIPS 2026: 2 paper reviewed

The lab

Deux Voir is a small, independent research identity exploring world models as memory substrates for multimodal reasoning. We are interested in the intersection of learned world models, latent representations, and language model reasoning. Our work is open-source and we welcome collaboration.