Souvik Kundu

Identifiers

• name variant Souvik Kundu 0.60 · backfill

Papers (15)

1. SPARQLe: Sub-Precision Activation Representation for Quantized LLM Inference cs.AR · 2026 · author #5
2. How Far Can Disaggregation Go? A Design-Space Exploration of Attention-FFN Disaggregation for Efficient MoE LLM Serving cs.LG · 2026 · author #7
3. Agentic AI Workload Characteristics cs.DC · 2026 · author #3
4. RankGuide: Tensor-Rank-Guided Routing and Steering for Efficient Reasoning cs.AI · 2026 · author #4
5. One Token Away from Collapse: The Fragility of Instruction-Tuned Helpfulness cs.CL · 2026 · author #3
6. SkipKV: Selective Skipping of KV Generation and Storage for Efficient Inference with Large Reasoning Models cs.AI · 2025 · author #10
7. Improving the Throughput of Diffusion-based Large Language Models via a Training-Free Confidence-Aware Calibration cs.LG · 2025 · author #7
8. Towards Understanding, Analyzing, and Optimizing Agentic AI Execution: A CPU-Centric Perspective cs.AI · 2025 · author #2
9. Top-H Decoding: Adapting the Creativity and Coherence with Bounded Entropy in Text Generation cs.CL · 2025 · author #3
10. MIST: A Co-Design Framework for Heterogeneous, Multi-Stage LLM Inference cs.AR · 2025 · author #5
11. Junk DNA Hypothesis: Pruning Small Pre-Trained Weights Irreversibly and Monotonically Impairs "Difficult" Downstream Tasks in LLMs cs.LG · 2023 · author #4
12. Exploiting Explicit Paths for Multi-hop Reading Comprehension cs.CL · 2018 · author #1
13. SpRRAM: A Predefined Sparsity Based Memristive Neuromorphic Circuit for Low Power Application cs.ET · 2018 · author #2
14. A Highly Parallel FPGA Implementation of Sparse Neural Network Training cs.DC · 2018 · author #4
15. A Question-Focused Multi-Factor Attention Network for Question Answering cs.CL · 2018 · author #1

Mentions

• 2606.00365 #5 · arxiv_oai · confidence 0.70 Souvik Kundu
• 2605.28302 #7 · arxiv_oai · confidence 0.70 Souvik Kundu
• 2605.26297 #3 · arxiv_oai · confidence 0.70 Souvik Kundu

Frequent Coauthors

• Massoud Pedram 4 shared papers
• Erfan Baghaei Potraghloo 3 shared papers
• Seyedarmin Azizi 3 shared papers
• Tushar Krishna 3 shared papers
• Abhimanyu Rajeshkumar Bambhaniya 2 shared papers
• Amir Yazdanbakhsh 2 shared papers
• Hanjiang Wu 2 shared papers
• Jiayi Tian 2 shared papers
• Madhu Kumar 2 shared papers
• Midhilesh Elavazhagan 2 shared papers
• Peter A. Beerel 2 shared papers
• Sharath Nittur Sridhar 2 shared papers
• Sudarshan Srinivasan 2 shared papers
• Suvinay Subramanian 2 shared papers
• Zhangyang Wang 2 shared papers
• Zheng Zhang 2 shared papers
• Aditya Akella 1 shared papers
• Ajay Jaiswal 1 shared papers
• Anand Raghunathan 1 shared papers
• Ankita Nayak 1 shared papers