The paper evaluates whether zero‑shot large language models (LLMs) can replace fine‑tuned natural language understanding (NLU) classifiers for intent detection in conversational systems. Experiments on ATIS and CLINC150 show that fine‑tuned RoBERTa outperforms Claude Haiku zero‑shot when ample in‑domain labels are available, but the LLM matches the supervised model on the broader 150‑intent CLINC150 schema. The authors identify three production‑relevant scenarios where LLMs excel—out‑of‑scope detection, robustness to ASR noise, and dynamic per‑deployment schemas—and present a decision framework to guide practitioners. whyItMatters:"The study provides concrete evidence and a practical framework for deciding when to deploy LLMs versus fine‑tuned models in real‑world conversational applications."
arXiv:2607. 27421v1 Announce Type: cross Abstract: Intent classification is a core component of task-oriented dialogue systems, yet practitioners have limited systematic guidance for selecting deployable open-weight language models under compute, latency, and robustness constraints.
DeepOpen is a decision‑oriented intent‑classification framework tailored for the BANKING77 and CLINC150 datasets. On BANKING77, a single English classifier trained with sharpness‑aware minimization (SAM) achieved a test accuracy of 94.1234% (2,899/3,080), the highest among three fixed SAM seeds. For CLINC150, a probability fusion of one Laya specialist and three DeBERTa‑v3‑large classifiers reached 98.0222% accuracy (4,411/4,500). The study reports strong task‑specific performance, notes selection effects, and highlights unresolved historical execution‑path differences rather than a new algorithmic advantage.
arXiv:2610.33813v1 Announce Type: new Abstract: Fine-tuning a language model for classification involves many design decisions. Comparisons of low-rank adaptation (LoRA) with full fine-tuning are com...
Large language models (LLMs) achieve strong relation extraction (RE), but their computational demands and reliance on proprietary APIs limit deployment in resource-constrained or privacy-sensitive settings. We investigate how far small language models (SLMs) can close this gap across general-domain and literary text.
The paper compares Complement Naive Bayes (NB) with zero‑shot and few‑shot large language models (LLMs) across a wide range of model sizes and text classification tasks. NB outperforms LLMs when labeled data is available, achieving comparable accuracy to large LLMs while running thousands of samples per second on a CPU. In zero‑data sentiment settings, LLMs still dominate, but NB remains the best choice for resource‑constrained HPC practitioners, and the authors provide a Kubernetes Helm operator to automate model selection.
The paper evaluates large language models (LLMs) as data quality annotators on two e-commerce tasks: entity matching and brand mislabeling. In entity matching, a simple rule-based baseline matched the LLM’s zero-shot performance (F1≈0.95), and a few-shot prompt actually lowered performance, highlighting the risk of small-sample prompt tuning. For brand mislabeling, the LLM outperformed a naive rule baseline (F1 0.833 vs 0.721) by leveraging background knowledge, and demonstrated high consistency across repeated runs (99.7% agreement).
arXiv:2606. 00467v1 Announce Type: cross Abstract: Large Language Models (LLMs) are increasingly used for zero-shot annotation and LLM-as-a-judge tasks, yet their reliability hinges on how model-internalized priors interact with user-provided instructions.
arXiv:2608. 10939v1 Announce Type: cross Abstract: Multilingual short-text classification supports operational systems such as content moderation, customer support routing, and intent recognition, yet aggregate evaluation often hides large differences between high-resource and low-resource languages.
The paper introduces a signed lexical gate that combines a sentence classifier’s logit margin with a sparse lexical model’s support for the predicted intent, assigning positive evidence to lexical agreement and negative evidence to a lexically favored competing intent. This gate retains more information than unsigned lexical confidence or a hard agreement rule and is calibrated via an independent binomial procedure to meet specified risk targets. Experiments on BANKING77, CLINC150, and HWU64 show that the proposed score reduces the area under the risk‑coverage curve by up to 15.8% and increases accepted coverage at low error rates, offering a compact, interpretable confidence enhancement for risk‑calibrated intent routing.
The paper introduces a multi‑signal pipeline for detecting hallucinations in large language models, combining fine‑tuned DeBERTa‑v3 classification, Monte Carlo Dropout uncertainty, and temperature‑scaled calibration. On the HaluEval benchmark it achieves high performance (F1 = 0.915, AUROC = 0.977) across QA, summarization, and dialogue, and shows that 25 % of training data yields 77 % of full‑data performance. The authors also demonstrate that applying Direct Preference Optimization to a Qwen2.5‑0.5B generator cuts hallucination rates from 85.5 % to 37.7 %, and that domain‑specific fine‑tuning (PubMedBERT on SciFact) outperforms general‑domain models for biomedical text.
AMLworld-Compact introduces an importance‑weighted downsampling protocol for evaluating large language models on AMLworld’s test splits, retaining all illicit transactions while sampling benign ones from supervised‑score strata. The HT‑Coreset method reduces evaluation targets by 271× on HI‑Small and 611× on LI‑Small, providing unbiased confusion‑count estimates with exact recall and ratio estimates for precision and F1. Experiments with boosted‑tree strata and Graph Feature Preprocessor features show lower F1 RMSE compared to uniform sampling, and audit tools reveal that lexical checks do not reliably indicate correctness of model outputs.
arXiv:2607. 14111v1 Announce Type: cross Abstract: Can small language models detect and report on perturbations their own internal activations?