arXiv subjects

Evan Kim

Publications and source records attributed to Evan Kim.

5 recordsLinked to original sources

arXiv2026

Turning Video Models into Generalist Robot Policies

Video generative models have emerged as a promising robotics backbone, capable of generating videos that depict the completion of complex tasks across embodiments and environments. Recent work proposes robot foundation models that jointly predict future observations and actions by finetuning video models with action-labeled data. In this paper, we test the limits of an alternative approach: leave the video planner as-is while training an embodiment-specific inverse dynamics model (IDM). This decoupling offers several natural benefits: the video planner remains embodiment-agnostic, different video models can be interchanged easily without re-training the IDM, and the IDM can be independently trained with readily available self-play data. We present a closed-loop, video-to-action policy that combines an action-free video world model with a carefully-designed IDM based on the robot embodiment Jacobian. We demonstrate that our IDM design is both data-efficient and scalable to high-dimensional action spaces. Our policy, which we coin the Video-to-Embodied Robot Action Model (VERA), achieves strong performance across simulated and real-world benchmarks, including zero-shot Panda arm manipulation and 16-DoF Allegro-hand dexterous cube re-orientation. The same video planner can be used across multiple embodiments by pairing it with different embodiment-specific IDMs. Our results show that decoupled video planning plus faithful video-to-action translation is a viable alternative route towards zero-shot, cross-embodiment, and generalizable robot control. More results are available on our project website: https://vera.csail.mit.edu.

cs.RO↗

arXiv2026

Scaling View Synthesis Transformers

Geometry-free view synthesis transformers have recently achieved state-of-the-art performance in Novel View Synthesis (NVS), outperforming traditional approaches that rely on explicit geometry modeling. Yet the factors governing their scaling with compute remain unclear. We present a systematic study of scaling laws for view synthesis transformers and derive design principles for training compute-optimal NVS models. Contrary to prior findings, we show that encoder-decoder architectures can be compute-optimal; we trace earlier negative results to suboptimal architectural choices and comparisons across unequal training compute budgets. Across several compute levels, we demonstrate that our encoder-decoder architecture, which we call the Scalable View Synthesis Model (SVSM), scales as effectively as decoder-only models, achieves a superior performance-compute Pareto frontier, and surpasses the previous state-of-the-art on real-world NVS benchmarks with substantially reduced training compute.

cs.CV↗

arXiv2022

ScGAN: A Generative Adversarial Network to Predict Hypothetical Superconductors

Despite having been discovered more than three decades ago, High Temperature Superconductors (HTSs) lack both an explanation for their mechanisms and a systematic way to search for them. To aid this search, this project proposes ScGAN, a Generative Adversarial Network (GAN) to efficiently predict new superconductors. ScGAN was trained on compounds in OQMD and then transfer learned onto the SuperCon database or a subset of it. Once trained, the GAN was used to predict superconducting candidates, and approximately 70\% of them were determined to be superconducting by a classification model--a 23-fold increase in discovery rate compared to manual search methods. Furthermore, more than 99\% of predictions were novel materials, demonstrating that ScGAN was able to potentially predict completely new superconductors, including several promising HTS candidates. This project presents a novel, efficient way to search for new superconductors, which may be used in technological applications or provide insight into the unsolved problem of high temperature superconductivity.

cond-mat.supr-con↗

arXiv2022

Replicating Portfolios: Constructing Permissionless Derivatives

The current design space of derivatives in Decentralized Finance (DeFi) relies heavily on oracle systems. Replicating market makers (RMMs) provide a mechanism for converting specific payoff functions to an associated Constant Function Market Makers (CFMMs). We leverage RMMs to replicate the approximate payoff of a Black-Scholes covered call option. RMM-01 is the first implementation of an on-chain expiring option mechanism that relies on arbitrage rather than an external oracle for price. We provide frameworks for derivative instruments and structured products achievable on-chain without relying on oracles. We construct long and binary options and briefly discuss perpetual covered call strategies commonly referred to as "theta vaults." Moreover, we introduce a procedure to eliminate liquidation risk in lending markets. The results suggest that CFMMs are essential for structured product design with minimized trust dependencies.

q-fin.CP↗

arXiv2020

A Comprehensive Evaluation of Multi-task Learning and Multi-task Pre-training on EHR Time-series Data

Multi-task learning (MTL) is a machine learning technique aiming to improve model performance by leveraging information across many tasks. It has been used extensively on various data modalities, including electronic health record (EHR) data. However, despite significant use on EHR data, there has been little systematic investigation of the utility of MTL across the diverse set of possible tasks and training schemes of interest in healthcare. In this work, we examine MTL across a battery of tasks on EHR time-series data. We find that while MTL does suffer from common negative transfer, we can realize significant gains via MTL pre-training combined with single-task fine-tuning. We demonstrate that these gains can be achieved in a task-independent manner and offer not only minor improvements under traditional learning, but also notable gains in a few-shot learning context, thereby suggesting this could be a scalable vehicle to offer improved performance in important healthcare contexts.

cs.LG↗