Ashish Goswami

Ashish Goswami, Satyam Kumar Modi, Santhosh Rishi Deshineni et al.

Text-to-image (T2I) generation has seen significant progress with diffusion models, enabling generation of photo-realistic images from text prompts. Despite this progress, existing methods still face challenges in following complex text prompts, especially those requiring compositional and multi-step reasoning. Given such complex instructions, SOTA models often make mistakes in faithfully modeling object attributes, and relationships among them. In this work, we present an alternate paradigm for T2I synthesis, decomposing the task of complex multi-step generation into three steps, (a) Generate: we first generate an image using existing diffusion models (b) Plan: we make use of Multi-Modal LLMs (MLLMs) to identify the mistakes in the generated image expressed in terms of individual objects and their properties, and produce a sequence of corrective steps required in the form of an edit-plan. (c) Edit: we make use of an existing text-guided image editing models to sequentially execute our edit-plan over the generated image to get the desired image which is faithful to the original instruction. Our approach derives its strength from the fact that it is modular in nature, is training free, and can be applied over any combination of image generation and editing models. As an added contribution, we also develop a model capable of compositional editing, which further helps improve the overall accuracy of our proposed approach. Our method flexibly trades inference time compute with performance on compositional text prompts. We perform extensive experimental evaluation across 3 benchmarks and 10 T2I models including DALLE-3 and the latest -- SD-3.5-Large. Our approach not only improves the performance of the SOTA models, by upto 3 points, it also reduces the performance gap between weaker and stronger models. $\href{https://dair-iitd.github.io/GraPE/}{https://dair-iitd.github.io/GraPE/}$

Ameer Shohail L, Ashish Goswami

Ultra-low latency services in 5G-Advanced demand deterministic delay and high-fidelity congestion signaling beyond peak throughput. While the Low Latency, Low Loss, Scalable Throughput (L4S) architecture enables sub-millisecond queuing through ECN-based feedback and Dual-Queue Coupled AQM, its integration within the 5G Core (5GC) remains functionally siloed. Current 3GPP Release 18/19 specifications provide mechanisms for L4S enablement, but they do not define a unified closed-loop framework that links application intent to verified service outcomes. To address this gap, we propose Closed-Loop L4S-as-a-Service (C-L4SaaS), an architectural framework that orchestrates the Network Exposure Function (NEF), Policy Control Function (PCF), and Network Data Analytics Function (NWDAF) for automated latency assurance. The framework translates high-level intent into enforceable PCC rules and uses NWDAF-driven compliance analytics, derived from User Plane Function (UPF) measurements, to trigger bounded policy adaptations. We model this interaction as a discrete-time feedback system and derive stability conditions and signaling overhead bounds to guide parameter selection under volatile wireless conditions. The proposed core-driven orchestration provides a standards-aligned path to expose, assure, and govern managed low-latency services in 5G-Advanced ecosystems.