Publications

Published

• The Energy of Falsehood: Detecting Hallucinations via Diffusion Model Likelihoods
  Arpit Singh Gautam, Kailash Talreja, Saurabh Jha
  FEVER Workshop, EACL 2026 · Published - ACL Anthology
  ACL Anthology Video + Poster
• Faithful Theory-of-Mind Distillation: Why Preference-Based Refinement Improves Imitation
  Arpit Singh Gautam, Saurabh Jha
  Theory-of-Mind Workshop (ToM4AI), AAAI 2026 · Published
  Paper (arXiv) Presentation
• Enhancing Lymphoma Detection Using Multi-Layer Hybrid Neural Networks
  Arpit Singh Gautam, et al.
  16th ICCCNT, IIT Indore (2025) · Oral presentation
• Real-Time QUIC Traffic Classification Using LightGBM with SHAP + LIME Explainability
  Arpit Singh Gautam, et al.
  16th ICCCNT, IIT Indore (2025)

Preprints

• RAMP: Reinforcement Adaptive Mixed-Precision Quantization for Efficient On-Device LLM Inference
  Arpit Singh Gautam, Saurabh Jha
  arXiv preprint (target: ICML) · Preprint on arXiv (2603.17891), targeting ICML
  arXiv:2603.17891
• CogniSQL-R1-Zero: Lightweight Reinforced Reasoning for Efficient SQL Generation
  Kushal Gajjar, Harshit Sikchi, Arpit Singh Gautam, Marc Hammons, Saurabh Jha
  arXiv preprint (+ released datasets) · Preprint on arXiv; paper + 2 datasets public
  arXiv:2507.06013
• StreamServe: Adaptive Speculative Flows for Low-Latency Disaggregated LLM Serving
  Satyam Kumar, Arpit Singh Gautam, Kailash Talreja, Saurabh Jha
  arXiv preprint · Preprint on arXiv (2604.09562)
  arXiv:2604.09562

Under review

• When AI-Discovered Scaling Laws Don't Always Transfer: A Calibration Case Study Across Open Model Families
  Arpit Singh Gautam, Swastik Vaish, Varun Tiwari, Hemlata Parmar
  EMNLP 2026 (under review) · Under review - decision Aug 2026
• Implicit World Models in Vision-Language-Action Policies and Why They Don't Stop the Robot from Doing the Wrong Thing
  Arpit Singh Gautam, Kushal Gajjar, Saanvi Singh
  CoRL 2026 (under review) · Under review - decision ~Sept 2026
• D-CLS: Training-Free Logit Suppression for Object Hallucination Mitigation in Vision-Language Models
  Arpit Singh Gautam, Swastik Vaish, Varun Tiwari, Jitendra Singh Yadav
  Under journal review · Under journal review

Earlier work

• Hyperspectral Image Classification with a Hybrid Deep Model
  Arpit Singh Gautam, et al.
  IIT-BHU research internship (unpublished) · Internship result (~91.9% accuracy); unpublished

Full list on Google Scholar.