collaborators

6 papers

cs.AI2026

MINCE: Shrinking LLM Evaluation Datasets via Few-Model Monte Carlo Calibration

Devleena Das, Rajeev Patwari, Vikram Kumar Bukka +3

Evaluating LLMs across many model variants -- quantized, fine-tuned, or deployment-specific -- requires running large benchmarks repeatedly, a process that can take tens of hours p…

cs.LG2026

Recover-LoRA for Aggressive Quantization: Reclaiming Accuracy in 2-Bit Language Models via Low-Rank Adaptation with Knowledge Distillation on Synthetic Data

Devleena Das, Rajeev Patwari, Elliott Delaye +1

Aggressive weight quantization to 2-bit precision offers substantial throughput and memory gains for large language model (LLM) inference, but typically incurs severe accuracy degr…

cs.LG2025

KV Pareto: Systems-Level Optimization of KV Cache and Model Compression for Long Context Inference

Sai Gokhale, Devleena Das, Rajeev Patwari +2

Long-context Large Language Models (LLMs) face significant memory bottlenecks during inference due to the linear growth of key-value (KV) cache with sequence length. While individu…

cs.CL2025

Recover-LoRA: Data-Free Accuracy Recovery of Degraded Language Models via Low-Rank Adaptation

Devleena Das, Rajeev Patwari, Ashish Sirasao

Inference optimizations such as quantization, pruning, format and datatype conversion, model export, and serialization can lead to functional degradations in language model task pe…

cs.CV2025

ByDeWay: Boost Your multimodal LLM with DEpth prompting in a Training-Free Way

Rajarshi Roy, Devleena Das, Ankesh Banerjee +3

We introduce ByDeWay, a training-free framework designed to enhance the performance of Multimodal Large Language Models (MLLMs). ByDeWay uses a novel prompting strategy called Laye…

cs.PF2025

Forecasting LLM Inference Performance via Hardware-Agnostic Analytical Modeling

Rajeev Patwari, Ashish Sirasao, Devleena Das

Large language models (LLMs) have been increasingly deployed as local agents on personal devices with CPUs, NPUs and integrated GPUs. However, forecasting inference performance on…