Author

Vineeth N. Balasubramanian

h-index 9 View on Semantic Scholar

We have 4 of 69 papers

We haven’t gathered this author’s papers yet. Follow them and we’ll fetch their work.

Not the right person? Other researchers publish under this name.

#small language model Preprint Oct 2026

When Visual Evidence Is Insufficient: How Instructions, Post-Training, and Internal Activations Shape Refusal in Multimodal Language Models

For reliable deployment, multimodal large language models (MLLMs) should recognize when an image provides insufficient evidence to answer a question and withhold unsupported answers. Yet refusal alone is not necessarily evidence of such recognition: models may refuse because of unrelated safety constraints or without i...

Ananth K. Suresh, Aveen Dayal, K. Gouthaman et al. · 0 citations

#artificial intelligence Preprint Sep 2026

ISA-Bench: A Benchmark for Computational Reasoning Across Instruction Set Architectures

A reasoning--execution gap (REG) analysis is introduced that reveals a recurring disconnect between identifying a plausible computational strategy and expressing it as a correct program in the target ISA.

A. Pola, Arkaprava Majumdar, Vineeth N. Balasubramanian · 0 citations

#artificial intelligence Preprint Sep 2026

Merge++: Universal Merge Refinement Through Data-Free Checkpoint Inversion

Merge++ is proposed, a post-hoc method that addresses model merging by inverting the expert checkpoints to synthesize task-representative images, then distilling expert knowledge into the merged model using those images.

A. Pola, Vineeth N. Balasubramanian · 0 citations

Preprint Aug 2026

Rethinking LLM Verification: Evidence Structure, Uncertainty, and Selective Refinement

A two-stage framework for medical hypothesis verification in multiple-choice settings that manages this tradeoff through targeted ontology grounding, applied only when the model abstains, and shows that abstention is not random but reflects genuine uncertainty, with abstained predictions associated with lower confidenc...

Uma S. Ranjan, Kunal Tilaganji, Aditya Koul et al. · 0 citations