Profiles2023-2026
Souvik Kundu
Intel Labs
13Papers
2Last author
25Unique co-authors
1First author
0Solo
First and last author come from byline position, which some venues assign alphabetically.
Recent papers13
- Improving the Throughput of Diffusion-based Large Language Models via a Training-Free Confidence-Aware CalibrationAuthor 7 of 7ACL Findings 2026
- RedVTP: Training-Free Acceleration of Diffusion Vision-Language Models Inference via Masked Token-Guided Visual Token PruningAuthor 5 of 6CVPR Findings 2026
- Top-H Decoding: Adapting the Creativity and Coherence with Bounded Entropy in Text GenerationAuthor 3 of 4NeurIPS 2025 · 13 citations
- MicroScopiQ: Accelerating Foundational Models through Outlier-Aware Microscaling QuantizationAuthor 2 of 3ISCA 2025 · 10 citations
- Ouromamba: a Data-Free Quantization Framework for Vision MambaAuthor 4 of 5ICCV 2025 · 3 citations
- Mitigating Hallucinations in Vision-Language Models through Image-Guided Head SuppressionAuthor 5 of 5EMNLP 2025 · 1 citation
- MambaExtend: A Training-Free Approach to Improve Long Context Extension of MambaAuthor 2 of 4ICLR 2025
- ShiftAddLLM: Accelerating Pretrained LLMs via Post-Training Multiplication-Less ReparameterizationAuthor 7 of 9NeurIPS 2024 · 47 citations
- Vision HGNN: An Image is More than a Graph of NodesAuthor 3 of 5ICCV 2023 · 86 citations
- SAL-ViT: Towards Latency Efficient Private Inference on ViT using Selective Attention Search with a Learnable Softmax ApproximationAuthor 3 of 5ICCV 2023 · 30 citations
- C2PI: An Efficient Crypto-Clear Two-Party Neural Network Private InferenceAuthor 3 of 6DAC 2023 · 7 citations
- Don't just prune by magnitude! Your mask topology is a secret weaponAuthor 2 of 4NeurIPS 2023 · 5 citations
- Learning to Linearize Deep Neural Networks for Secure and Efficient Private InferenceAuthor 1 of 5ICLR 2023 · 3 citations
