V. S. Raghu Parupudi

V. S. Raghu Parupudi, Harsha Ponnada, Aditi Kaushal et al.

Reference-free evaluation of large language model (LLM) creativity relies on perplexity, entropy, and top-1 margin. We show that a much stronger signal lives one step earlier in the pipeline: in how sampling temperature \emph{reshapes} the model's token distribution before the next token is drawn. On Llama-3.1-8B-Instruct generations of 500 open-ended creative prompts at $T \in \{0.3, 0.8, 1.5\}$, a single per-token feature derived from this reshaping predicts the within-prompt creativity rank at Spearman $ρ{=}0.918$ against an averaged gpt-4o\,/\,gemini-2.5-pro judge ($n{=}500$) and $ρ{=}0.870$ against a three-rater human-majority ranking ($n{=}150$). Each of four standard reference-free baselines (self-perplexity, mean predictive entropy, top-1 margin, gzip compression ratio) tops out at $|ρ|\!\approx\!0.76$ on both ground truths: a gap of $+0.165$ on averaged-LLM and $+0.110$ on human-majority, both far larger than the spread among the baselines themselves. The two ground-truth panels agree with each other at $ρ{=}0.83$, above the inter-human ceiling of $ρ{=}0.77$, so the comparison is not bottlenecked by judge noise. Mechanistically, the win comes from a sharp distributional signature of the incoherence regime: at $T{=}1.5$ the cumulative-mass width $n_{95}(q)$ inflates from $\sim\!1$ to ${\sim}\!131$ tokens and post-temperature mass leaks off the pre-temperature top-$90\%$ plausible set by about $13$ percentage points. The per-token aggregates do not separate $T{=}0.8$ from $T{=}0.3$; discriminating the two coherent regimes is left to sequence-level features.

V. S. Raghu Parupudi

Reference-free metrics like self-perplexity are strongly biased against creative text generation. We propose the Confidence Score (CS), derived from a model's output probability distribution, as a less biased alternative. Experiments on gpt-4o-mini show that while fluency-based metrics prefer novel responses in 0\% of cases on 99 creative prompts, our CS does so 19% of the time, a statistically significant difference (95% CI for difference: [11.1%, 27.3%]). We also show that CS effectively distinguishes between easy, medium, and hard tasks, confirmed by non-overlapping confidence intervals. The Confidence Score thus mitigates the creativity bias of traditional metrics while retaining their core evaluative strengths, offering a more balanced assessment for modern LLMs.

V. S. Raghu Parupudi

A central question in artificial intelligence is the extent to which machine learning models comprehend mathematics. To address this, we propose a novel framework for measuring mathematical reasoning that moves beyond standard benchmarks to diagnose specific failure points. Our method first generates structured, step-by-step reasoning from gpt-3.5-turbo on the GSM8K dataset. We then use a more capable analyst model, gpt-4o-mini, to categorize errors and, crucially, perform an unsupervised clustering of every reasoning sentence to identify emergent "reasoning modes." This analysis reveals a cognitive profile with a stark, nonhuman-like brittleness: while the model achieves near-perfect accuracy on procedural modes like sequential calculation, its performance on modes requiring combinatorial reasoning with restrictions plummets. By identifying and quantifying the reliability of these distinct reasoning skills, our work provides a more granular method to evaluate mathematical comprehension and offers a precise roadmap for developing new capabilities and more reliable future applications.

V. S. Raghu Parupudi

Vector comparison in high dimensions is a fundamental task in NLP, yet it is dominated by two baselines: the raw dot product, which is unbounded and sensitive to vector norms, and the cosine similarity, which discards magnitude information entirely. This paper challenges both standards by proposing and rigorously evaluating a new class of parameter-free, magnitude-aware similarity metrics. I introduce two such functions, Overlap Similarity (OS) and Hyperbolic Tangent Similarity (HTS), designed to integrate vector magnitude and alignment in a more principled manner. To ensure that my findings are robust and generalizable, I conducted a comprehensive evaluation using four state-of-the-art sentence embedding models (all-MiniLM-L6-v2, all-mpnet-base-v2, paraphrase-mpnet-base-v2, and BAAI/bge-large-en-v1.5) across a diverse suite of eight standard NLP benchmarks, including STS-B, SICK, Quora, and PAWS. Using the Wilcoxon signed-rank test for statistical significance, my results are definitive: on the tasks requiring holistic semantic understanding (paraphrase and inference), both OS and HTS provide a statistically significant improvement in Mean Squared Error over both the raw dot product and cosine similarity, regardless of the underlying embedding model.Crucially, my findings delineate the specific domain of advantage for these metrics: for tasks requiring holistic semantic understanding like paraphrase and inference, my magnitude-aware metrics offer a statistically superior alternative. This significant improvement was not observed on benchmarks designed to test highly nuanced compositional semantics (SICK, STS-B), identifying the challenge of representing compositional text as a distinct and important direction for future work.