arXiv subjects

Hari Bandi

Publications and source records attributed to Hari Bandi.

3 recordsLinked to original sources

arXiv2026

From Attribution to Abstention: Training-Free Attention-Based Auditing for Clinical Summarization

Deploying multimodal large language models (MLLMs) for clinical summarization demands not only fluent generation but also transparency about where each statement originates-and a mechanism to flag when statements lack evidential support. We present ClinTrace, a training-free framework that extracts two clinically useful signals from the decoder attention weights that every transformer-based MLLM already produces during generation: (i) fine-grained source attributions linking each output sentence to supporting text spans or images, and (ii) per-sentence groundedness scores that identify poorly supported claims as candidate hallucinations. Both signals are derived from the same attention tensors in a single pass, requiring no retraining, no auxiliary models, and no additional inference cost. We evaluate on two clinical summarization tasks: doctor-patient dialogue summarization (CliConSummation) and radiology report summarization (MIMIC-CXR) using a general-purpose MLLM (Qwen3-8B) and a medical-finetuned model (HuatuoGPT-Vision-7B). For source attribution, ClinTrace achieves over 92% text F1 on radiology and 88% on dialogue summarization, substantially outperforming embedding-based and self-attribution baselines. For hallucination detection, groundedness scores achieve 0.77 AUROC with the medical-finetuned model: competitive with embedding-based confidence at zero additional cost, and enable an abstention mechanism that improves faithfulness from 61.7% to 72.6% by withholding the least: grounded 20% of output for clinician review. Notably, medical finetuning substantially improves the reliability of attention-based hallucination detection, suggesting that domain adaptation produces more semantically structured attention patterns amenable to self-auditing.

cs.CL↗

arXiv2024

Debate, Deliberate, Decide (D3): A Cost-Aware Adversarial Framework for Reliable and Interpretable LLM Evaluation

The evaluation of Large Language Models (LLMs) remains challenging due to inconsistency, bias, and the absence of transparent decision criteria in automated judging. We present Debate, Deliberate, Decide (D3), a cost-aware, adversarial multi-agent framework that orchestrates structured debate among role-specialized agents (advocates, a judge, and an optional jury) to produce reliable and interpretable evaluations. D3 instantiates two complementary protocols: (1) Multi-Advocate One-Round Evaluation (MORE), which elicits k parallel defenses per answer to amplify signal via diverse advocacy, and (2) Single-Advocate Multi-Round Evaluation (SAMRE) with budgeted stopping, which iteratively refines arguments under an explicit token budget and convergence checks. We develop a probabilistic model of score gaps that (i) characterizes reliability and convergence under iterative debate and (ii) explains the separation gains from parallel advocacy. Under mild assumptions, the posterior distribution of the round-r gap concentrates around the true difference and the probability of mis-ranking vanishes; moreover, aggregating across k advocates provably increases expected score separation. We complement theory with a rigorous experimental suite across MT-Bench, AlignBench, and AUTO-J, showing state-of-the-art agreement with human judgments (accuracy and Cohen's kappa), reduced positional and verbosity biases via anonymization and role diversification, and a favorable cost-accuracy frontier enabled by budgeted stopping. Ablations and qualitative analyses isolate the contributions of debate, aggregation, and anonymity. Together, these results establish D3 as a principled, practical recipe for reliable, interpretable, and cost-aware LLM evaluation.

cs.CL↗

arXiv2021

The Price of Diversity

Systemic bias with respect to gender, race and ethnicity, often unconscious, is prevalent in datasets involving choices among individuals. Consequently, society has found it challenging to alleviate bias and achieve diversity in a way that maintains meritocracy in such settings. We propose (a) a novel optimization approach based on optimally flipping outcome labels and training classification models simultaneously to discover changes to be made in the selection process so as to achieve diversity without significantly affecting meritocracy, and (b) a novel implementation tool employing optimal classification trees to provide insights on which attributes of individuals lead to flipping of their labels, and to help make changes in the current selection processes in a manner understandable by human decision makers. We present case studies on three real-world datasets consisting of parole, admissions to the bar and lending decisions, and demonstrate that the price of diversity is low and sometimes negative, that is we can modify our selection processes in a way that enhances diversity without affecting meritocracy significantly, and sometimes improving it.

cs.CY↗