arXiv subjects

Armin Gerami

Publications and source records attributed to Armin Gerami.

9 recordsLinked to original sources

arXiv2026

Same Question, Different Answers: Evaluating LLM Reliability Beyond Accuracy

Large language models (LLMs) often achieve strong accuracy on benchmarks, yet it remains unclear how reliably they apply this knowledge when the same question is phrased in different but equivalent ways. In this work, we study how model answers change under meaning-preserving paraphrases across factual question answering and mathematical reasoning tasks. Across four benchmarks and 13 models, we find that model outputs frequently depend on the exact wording of the prompt. While overall accuracy typically changes only modestly across paraphrases, instance-level behavior is far less stable: for many questions, models alternate between correct and incorrect answers depending on phrasing, with mismatch rates reaching more than 23%. Conditioning on questions that are answered correctly in their original form reveals even larger failures measured by answer flip rates, showing that single-prompt correctness is often a poor indicator of reliability. At the same time, we find that models often produce a correct answer for at least one paraphrase of a question, suggesting that the underlying knowledge is present but inconsistently retrieved. Building on this observation, we show that a simple self-paraphrasing strategy can partially recover this latent knowledge and improve performance at inference time. Together, these findings suggest that standard accuracy metrics can mask substantial instability, and that evaluating consistency across equivalent inputs provides a clearer picture of LLM reliability.

cs.AI↗

arXiv2026

On The Application of Linear Attention in Multimodal Transformers

Multimodal Transformers serve as the backbone for state-of-the-art vision-language models, yet their quadratic attention complexity remains a critical barrier to scalability. In this work, we investigate the viability of Linear Attention (LA) as a high-efficiency alternative within multimodal frameworks. By integrating LA, we reduce the computational overhead from quadratic to linear relative to sequence length while preserving competitive performance. We evaluate our approach across ViT-S/16, ViT-B/16, and ViT-L/16 architectures trained on the LAION-400M dataset, with validation focused on ImageNet-21K zero-shot accuracy. Our systematic evaluation demonstrates that Linear Attention not only yields significant computational savings but also adheres to the same scaling laws as standard softmax attention. These findings position Linear Attention as a robust, scalable solution for next-generation multimodal Transformers tasked with processing increasingly large and complex datasets.

cs.CV↗

arXiv2025

Quantifying Document Impact in RAG-LLMs

Retrieval Augmented Generation (RAG) enhances Large Language Models (LLMs) by connecting them to external knowledge, improving accuracy and reducing outdated information. However, this introduces challenges such as factual inconsistencies, source conflicts, bias propagation, and security vulnerabilities, which undermine the trustworthiness of RAG systems. A key gap in current RAG evaluation is the lack of a metric to quantify the contribution of individual retrieved documents to the final output. To address this, we introduce the Influence Score (IS), a novel metric based on Partial Information Decomposition that measures the impact of each retrieved document on the generated response. We validate IS through two experiments. First, a poison attack simulation across three datasets demonstrates that IS correctly identifies the malicious document as the most influential in $86\%$ of cases. Second, an ablation study shows that a response generated using only the top-ranked documents by IS is consistently judged more similar to the original response than one generated from the remaining documents. These results confirm the efficacy of IS in isolating and quantifying document influence, offering a valuable tool for improving the transparency and reliability of RAG systems.

cs.IR↗

arXiv2025

Transformer Based Linear Attention with Optimized GPU Kernel Implementation

The original softmax-based attention mechanism (regular attention) in the extremely successful Transformer architecture computes attention between $N$ tokens, each embedded in a $D$-dimensional head, with a time complexity of $O(N^2D)$. Given the success of Transformers, improving their runtime during both training and inference is a popular research area. One such approach is the introduction of the linear attention (LA) mechanisms, which offers a linear time complexity of $O(ND^2)$ and have demonstrated comparable accuracy to regular attention. However, LA in practice lags behind its theoretical efficiency. We propose a novel method for LA's forward and backward passes, along with a highly-optimized CUDA implementation. Our approach outperforms the state-of-the-art by 3.3 times in speed and reduces memory consumption by 3.6 times. We validate these improvements in both single-layer and end-to-end settings by training a 1.4 billion parameter language model, which demonstrates similar expressivity to regular attention on major reasoning benchmarks.

cs.LG↗

arXiv2025

A Scalable MVDR Beamforming Algorithm That is Linear in the Number of Antennas

The Minimum Variance Distortionless Response (MVDR) beamforming technique is widely applied in array systems to mitigate interference. However, applying MVDR to large arrays is computationally challenging; its computational complexity scales cubically with the number of antenna elements. In this paper, we introduce a scalable MVDR beamforming method tailored for massive arrays. Our approach, which is specific to scenarios where the signal of interest is below the noise floor (e.g.,~GPS), leverages the Sherman-Morrison formula, low-rank Singular Value Decomposition (SVD) approximations, and algebraic manipulation. Using our approach, we reduce the computational complexity from cubic to linear in the number of antennas. We evaluate the proposed method through simulations, comparing its computational efficiency and beamforming accuracy with the conventional MVDR approach. Our method significantly reduces the computational load while maintaining high beamforming accuracy for large-scale arrays. This solution holds promise for real-time applications of MVDR beamforming in fields like radar, sonar, and wireless communications, where massive antenna arrays are proliferating.

eess.SP↗

arXiv2025

Auditing Algorithmic Bias in Transformer-Based Trading

Transformer models have become increasingly popular in financial applications, yet their potential risk making and biases remain under-explored. The purpose of this work is to audit the reliance of the model on volatile data for decision-making, and quantify how the frequency of price movements affects the model's prediction confidence. We employ a transformer model for prediction, and introduce a metric based on Partial Information Decomposition (PID) to measure the influence of each asset on the model's decision making. Our analysis reveals two key observations: first, the model disregards data volatility entirely, and second, it is biased toward data with lower-frequency price movements.

cs.LG↗

arXiv2025

Room Impulse Response Synthesis via Differentiable Feedback Delay Networks for Efficient Spatial Audio Rendering

We introduce a computationally efficient and tunable feedback delay network (FDN) architecture for real-time room impulse response (RIR) rendering that addresses the computational and latency challenges inherent in traditional convolution and Fourier transform based methods. Our approach directly optimizes FDN parameters to match target RIR acoustic and psychoacoustic metrics such as clarity and definition through novel differentiable programming-based optimization. Our method enables dynamic, real-time adjustments of room impulse responses that accommodates listener and source movement. When combined with previous work on representation of head-related impulse responses via infinite impulse responses, an efficient rendering of auditory objects is possible when the HRIR and RIR are known. Our method produces renderings with quality similar to convolution with long binaural room impulse response (BRIR) filters, but at a fraction of the computational cost.

eess.AS↗

arXiv2024

GUST: Graph Edge-Coloring Utilization for Accelerating Sparse Matrix Vector Multiplication

Sparse matrix-vector multiplication (SpMV) plays a vital role in various scientific and engineering fields, from scientific computing to machine learning. Traditional general-purpose processors often fall short of their peak performance with sparse data, leading to the development of domain-specific architectures to enhance SpMV. Yet, these specialized approaches, whether tailored explicitly for SpMV or adapted from matrix-matrix multiplication accelerators, still face challenges in fully utilizing hardware resources as a result of sparsity. To tackle this problem, we introduce GUST, a hardware/software co-design, the key insight of which lies in separating multipliers and adders in the hardware, thereby enabling resource sharing across multiple rows and columns, leading to efficient hardware utilization and ameliorating negative performance impacts from sparsity. Resource sharing, however, can lead to collisions, a problem we address through a specially devised edge-coloring scheduling algorithm. Our comparisons with various prior domain specific architectures using real-world datasets shows the effectiveness of GUST, with an average hardware utilization of $33.67\%$.

cs.AR↗

arXiv2024

FAST: Factorizable Attention for Speeding up Transformers

Motivated by the factorization inherent in the original fast multipole method and the improved fast Gauss transform we introduce a factorable form of attention that operates efficiently in high dimensions. This approach reduces the computational and memory complexity of the attention mechanism in transformers from $O(N^2)$ to $O(N)$. In comparison to previous attempts, our work presents a linearly scaled attention mechanism that maintains the full representation of the attention matrix without compromising on sparsification and incorporates the all-to-all relationship between tokens. We explore the properties of our new attention metric and conduct tests in various standard settings. Results indicate that our attention mechanism has a robust performance and holds significant promise for diverse applications where self-attention is used.

cs.LG↗