Neel Joshi
YOU?
Author Swipe
View article: Programming cellular condensates for living materials using an intrinsically disordered protein display platform
Programming cellular condensates for living materials using an intrinsically disordered protein display platform Open
Self-organization underpins the emergence of complex structure in living systems but remains a major challenge for engineering synthetic multicellular materials. Here, we present intrinsically disordered protein display platform (iDP 2 ), …
View article: What MLLMs Learn about When they Learn about Multimodal Reasoning: Perception, Reasoning, or their Integration?
What MLLMs Learn about When they Learn about Multimodal Reasoning: Perception, Reasoning, or their Integration? Open
Multimodal reasoning models have recently shown promise on challenging domains such as olympiad-level geometry, yet their evaluation remains dominated by aggregate accuracy, a single score that obscures where and how models are improving. …
View article: Phi-4-reasoning Technical Report
Phi-4-reasoning Technical Report Open
We introduce Phi-4-reasoning, a 14-billion parameter reasoning model that achieves strong performance on complex reasoning tasks. Trained via supervised fine-tuning of Phi-4 on carefully curated set of "teachable" prompts-selected for the …
View article: Inference-Time Scaling for Complex Tasks: Where We Stand and What Lies Ahead
Inference-Time Scaling for Complex Tasks: Where We Stand and What Lies Ahead Open
Inference-time scaling can enhance the reasoning capabilities of large language models (LLMs) on complex problems that benefit from step-by-step problem solving. Although lengthening generated scratchpads has proven effective for mathemati…
View article: Programming cellular condensates for living materials using an intrinsically disordered protein display platform (iDP<sup>2</sup>)
Programming cellular condensates for living materials using an intrinsically disordered protein display platform (iDP<sup>2</sup>) Open
Self-organization of multicellular systems is vital for building structure in living system but remains underexplored in engineered living materials. We developed iDP 2 , a platform enabling high-density display of intrinsically disordered…
View article: MM-GEN: Enhancing Task Performance Through Targeted Multimodal Data Curation
MM-GEN: Enhancing Task Performance Through Targeted Multimodal Data Curation Open
Vision-language models (VLMs) are highly effective but often underperform on specialized tasks; for example, Llava-1.5 struggles with chart and diagram understanding due to scarce task-specific training data. Existing training data, source…
View article: Bioengineering approach for the design of magnetic bacterial cellulose membranes
Bioengineering approach for the design of magnetic bacterial cellulose membranes Open
Biopolymer research has led to the development of novel products through innovative strategies. Their functionalization is typically achieved by physical/chemical methods that require harsh chemicals or mechanical treatments. These functio…
View article: In situ neutralization of the genotoxicity of pathogenic gut bacteria with the polyketide synthase island
In situ neutralization of the genotoxicity of pathogenic gut bacteria with the polyketide synthase island Open
The gut microbiota produce beneficial and deleterious metabolites to influence host physiology, with colibactin from pks + bacteria one of the most studied metabolites associated with host genotoxicity and tumorigenesis. However, no FDA-…
View article: BenchAgents: Multi-Agent Systems for Structured Benchmark Creation
BenchAgents: Multi-Agent Systems for Structured Benchmark Creation Open
Evaluation insights are limited by the availability of high-quality benchmarks. As models evolve, there is a need to create benchmarks that can measure progress on new and complex generative capabilities. However, manually creating new ben…
View article: Unearthing Skill-Level Insights for Understanding Trade-Offs of Foundation Models
Unearthing Skill-Level Insights for Understanding Trade-Offs of Foundation Models Open
With models getting stronger, evaluations have grown more complex, testing multiple skills in one benchmark and even in the same instance at once. However, skill-wise performance is obscured when inspecting aggregate accuracy, under-utiliz…
View article: Eureka: Evaluating and Understanding Large Foundation Models
Eureka: Evaluating and Understanding Large Foundation Models Open
Rigorous and reproducible evaluation is critical for assessing the state of the art and for guiding scientific advances in Artificial Intelligence. Evaluation is challenging in practice due to several reasons, including benchmark saturatio…
View article: A Raspberry Pi-Based Text Reader & Object Detection System
A Raspberry Pi-Based Text Reader & Object Detection System Open
This research introduces a novel application of the raspberry pi platform in the development of an integrated text reader and object detection system. The system aims to assist visually impaired individuals and enhance overall accessibilit…
View article: Is A Picture Worth A Thousand Words? Delving Into Spatial Reasoning for Vision Language Models
Is A Picture Worth A Thousand Words? Delving Into Spatial Reasoning for Vision Language Models Open
Large language models (LLMs) and vision-language models (VLMs) have demonstrated remarkable performance across a wide range of tasks and domains. Despite this promise, spatial understanding and reasoning -- a fundamental component of human…
View article: Mechanically Tunable, Compostable, Healable and ScalableEngineered Living Materials
Mechanically Tunable, Compostable, Healable and ScalableEngineered Living Materials Open
Novel design strategies are essential to realize the full potential of Engineered Living Materials (ELMs), including their biodegradability, manufacturability, sustainability, and ability to tailor functional properties. Toward these goals…
View article: Mechanically Tunable, Compostable, Healable and Scalable Engineered Living Materials
Mechanically Tunable, Compostable, Healable and Scalable Engineered Living Materials Open
Novel design strategies are essential to realize the full potential of Engineered Living Materials (ELMs), including their biodegradability, manufacturability, sustainability, and ability to tailor functional properties. Toward these goals…
View article: Optimizing University Course Timetabling Using Constraint Satisfaction Models
Optimizing University Course Timetabling Using Constraint Satisfaction Models Open
The giving out of the course timetable is a problem found at ever increasing complexity with the growth of universities. This problem demands the prompt allocation of courses along with instructors, rooms and times while satisfying a numbe…
View article: Microbially Driven Reversible Actuation and Color Changing Materials
Microbially Driven Reversible Actuation and Color Changing Materials Open
A common feature of natural living systems that is underexplored in the field of engineered living materials (ELMs) is macroscale mechanical actuation, as driven by active cellular processes. Here we demonstrate an ELM wherein Escherichia …
View article: HoloAssist: an Egocentric Human Interaction Dataset for Interactive AI Assistants in the Real World
HoloAssist: an Egocentric Human Interaction Dataset for Interactive AI Assistants in the Real World Open
Building an interactive AI assistant that can perceive, reason, and collaborate with humans in the real world has been a long-standing pursuit in the AI community. This work is part of a broader research effort to develop intelligent agent…
View article: Beyond Generation: Harnessing Text to Image Models for Object Detection and Segmentation
Beyond Generation: Harnessing Text to Image Models for Object Detection and Segmentation Open
We propose a new paradigm to automatically generate training data with accurate labels at scale using the text-to-image synthesis frameworks (e.g., DALL-E, Stable Diffusion, etc.). The proposed approach1 decouples training data generation …
View article: Controllable Text-to-Image Generation with GPT-4
Controllable Text-to-Image Generation with GPT-4 Open
Current text-to-image generation models often struggle to follow textual instructions, especially the ones requiring spatial reasoning. On the other hand, Large Language Models (LLMs), such as GPT-4, have shown remarkable precision in gene…
View article: Periplasmic stress contributes to a tradeoff between protein secretion and cell growth in E. Coli Nissile
Periplasmic stress contributes to a tradeoff between protein secretion and cell growth in E. Coli Nissile Open
Maximizing protein secretion is an important target in the design of engineered living systems. In this paper, we characterize a tradeoff between cell growth and per cell protein secretion in the curli biofilm secretion system of E Coli Ni…
View article: Periplasmic stress contributes to a trade-off between protein secretion and cell growth in <i>Escherichia coli</i> Nissle 1917
Periplasmic stress contributes to a trade-off between protein secretion and cell growth in <i>Escherichia coli</i> Nissle 1917 Open
Maximizing protein secretion is an important target in the design of engineered living systems. In this paper, we characterize a trade-off between cell growth and per-cell protein secretion in the curli biofilm secretion system of Escheric…
View article: PatchBlender: A Motion Prior for Video Transformers
PatchBlender: A Motion Prior for Video Transformers Open
Transformers have become one of the dominant architectures in the field of computer vision. However, there are yet several challenges when applying such architectures to video data. Most notably, these models struggle to model the temporal…
View article: Single domain antibodies against enteric pathogen virulence factors are active as curli fiber fusions on probiotic E. coli Nissle 1917
Single domain antibodies against enteric pathogen virulence factors are active as curli fiber fusions on probiotic E. coli Nissle 1917 Open
Enteric microbial pathogens, including Escherichia coli , Shigella and Cryptosporidium species, take a particularly heavy toll in low-income countries and are highly associated with infant mortality. We describe here a means to display ant…
View article: Neural-Sim: Learning to Generate Training Data with NeRF
Neural-Sim: Learning to Generate Training Data with NeRF Open
Training computer vision models usually requires collecting and labeling vast amounts of imagery under a diverse set of scene configurations and properties. This process is incredibly time-consuming, and it is challenging to ensure that th…