publications

Research on information retrieval, learned representations, and efficient adaptation.

  1. ReNeuIR @ SIGIR
    Most MaxSim Winners Flip, Retrieval Survives: Low-Margin Substitution in Sign-Coded Late Interaction
    Anirudh Lakhotia and Nischal Helagally Shantharaju
    5th Workshop on Reaching Efficiency in Neural Information Retrieval (ReNeuIR), co-located with ACM SIGIR, 2026
    Extreme compression changes most token-level matches, yet retrieval remains robust through low-margin substitutions.
  2. ICEENG
    Analysis of Sampling Strategies for Multi-Task Learning in Transformer Models
    Anirudh Lakhotia*, Akash Kamalesh*, Nischal Helagally Shantharaju*, Prerana Sanjay Kulkarni*, and Gowri Srinivasa*
    International Conference on Electrical Engineering (ICEENG), 2025
    *Equal contribution.
    Task-sampling strategies shape the internal representations learned by transformers.
  3. FITML @ NeurIPS
    UnoLoRA: Single Low-Rank Adaptation for Efficient Multitask Fine-tuning
    Akash Kamalesh*, Anirudh Lakhotia*, Nischal Helagally Shantharaju*, Prerana Sanjay Kulkarni*, and Gowri Srinivasa
    Workshop on Fine-Tuning in Machine Learning, co-located with NeurIPS, 2024
    *Equal contribution among the first four authors.
    A single shared low-rank adapter supports efficient fine-tuning across multiple tasks.