Related papers: MarginGate: Sparse Margin-Triggered Verification f…
While random demonstration labels barely hurt in-context learning (Min et al., 2022), we show that homogeneous labels--even semantically valid ones--collapse accuracy to <=12% across six models (Pythia, Llama, Qwen; 0.8B--8B) and four…
Reasoning models often outperform smaller models but at 3--5$\times$ higher cost and added latency. We present entropy-guided refinement: a lightweight, test-time loop that uses token-level uncertainty to trigger a single, targeted…
Gating mechanisms have been widely utilized, from early models like LSTMs and Highway Networks to recent state space models, linear attention, and also softmax attention. Yet, existing literature rarely examines the specific effects of…
This paper addresses the critical challenge of deriving interpretable confidence scores from generative language models (LLMs) when applied to multi-label content safety classification. While models like LLaMA Guard are effective for…
Deploying LLMs raises two coupled challenges: (1) monitoring--estimating where a model underperforms as traffic and domains drift--and (2) improvement--prioritizing data acquisition to close the largest performance gaps. We test whether an…
We introduce obfuscation testing, a novel methodology for validating whether large language models detect structural market patterns through causal reasoning rather than temporal association. Testing three dealer hedging constraint patterns…
Detection of level shifts in a noisy signal, or trend break detection, is a problem that appears in several research fields, from biophysics to optics and economics. Although many algorithms have been developed to deal with such problem,…
Large language models (LLMs) are increasingly used for high-stakes decisions, yet their susceptibility to spurious features remains poorly characterized. We introduce ICE-Guard, a framework applying intervention consistency testing to…
Large language models (LLMs) have demonstrated state-of-the-art performance across various tasks. However, the latency of inference and the large GPU memory consumption of LLMs restrict their deployment performance. Recently, there have…
Existing benchmarks for tool-augmented language models (TaLMs) lack fine-grained control over task difficulty and remain vulnerable to data contamination. We present FuncBenchGen, a unified, contamination-free framework that evaluates TaLMs…
Auto-regressive decoding in Large Language Models (LLMs) is inherently memory-bound: every generation step requires loading the model weights and intermediate results from memory (e.g., High-Bandwidth Memory (HBM) for GPU servers), making…
Large Language Models (LLMs) often exhibit strong linguistic abilities while remaining unreliable on multi-step reasoning tasks, particularly when deployed without additional training or fine-tuning. In this work, we study inference-time…
Experiments conducted on open-access cloud-based IBM Quantum devices are presented for characterizing their fault tolerance using $[4,2,2]$-encoded gate sequences. Up to 100 logical gates are activated in the IBMQ Bogota and IBMQ Santiago…
We present a universal scheme of pulsed operations for the IBM oscillator-stabilized flux qubit comprising the CPHASE gate, single-qubit preparations and measurements. Based on numerical simulations, we argue that the error rates for these…
We present a practical system for privacy-aware large language model (LLM) inference that splits a transformer between a trusted local GPU and an untrusted cloud GPU, communicating only intermediate activations over the network. Our system…
Coverage closure is the most time-consuming phase of hardware verification, and recent large language model (LLM)-based coding agents offer a promising approach to automated stimulus generation. However, prior LLM-based flows do not…
This thesis investigates the physics performance, trigger efficiency, and Field Programmable Gate Array (FPGA) implementation of machine learning (ML)-based algorithms for Lorentz-boosted $H\rightarrow b\bar{b}$ tagging within the CMS…
The computational cost of training multimodal large language models (MLLMs) grows rapidly with the number of processed tokens. Existing efficiency methods mainly target inference via token reduction or merging, offering limited benefits…
Efficient and realistic error decoding is crucial for fault-tolerant quantum computation (FTQC) on near-term devices. While decoding is a classical post-processing task, its effectiveness depends on accurately modeling quantum noise, which…
Large Language Models (LLMs) often rely on test-time scaling via parallel decoding (for example, 512 samples) to boost reasoning accuracy, but this incurs substantial compute. We introduce CoRefine, a confidence-guided self-refinement…