LoRA, PEFT, instruction tuning and domain adaptation — adapting a pretrained model without paying to train one.
arXiv:2602.08062v2 Announce Type: replace Abstract: Large Language Models have demonstrated remarkable capabilities in natural language understanding. However, they remain susceptible to malicious pr...
arXiv:2610.26986v1 Announce Type: new Abstract: Contextual automatic speech recognition (ASR) improves rare-entity recognition by supplying candidate hotwords, but retrieved candidates may be incompl...
arXiv:2610.27237v1 Announce Type: new Abstract: We present a study on controlled automatic simplification of Russian texts using preference-based optimization algorithms with a focus on adapting text...
arXiv:2605.22140v3 Announce Type: replace Abstract: Psychological counseling unfolds over many sessions, and what a client brings to each session depends on what has happened in their life since the...
arXiv:2410.18362v4 Announce Type: replace-cross Abstract: Web development involves turning UI designs into functional webpages, which can be difficult for both beginners and experienced developers du...
arXiv:2610.26933v1 Announce Type: new Abstract: An event camera emits events at a rate given by the scene motion, object texture, illumination, its own contrast thresholds, and other settings. Across...
arXiv:2610.27656v1 Announce Type: new Abstract: Medical visual grounding requires models to align clinical descriptions with spatial evidence across heterogeneous imaging domains. However, existing m...
arXiv:2610.28382v1 Announce Type: new Abstract: Training multimodal search agents requires tasks that make visually grounded search necessary for determining the answer. Textual identity clues and un...
arXiv:2610.28625v1 Announce Type: new Abstract: Understanding human needs and selecting responses require distinct judgments that aggregate accuracy can obscure. We introduce PsycheMME, a video bench...
arXiv:2607.12782v2 Announce Type: replace Abstract: Hyperspectral foundation models learn transferable spectral-spatial representations from large-scale unlabeled data. They provide an effective para...
arXiv:2610.26820v2 Announce Type: replace Abstract: A geometric style learned from one family of shapes is typically applied to new shapes from the same family. We consider a more challenging setting...
The study evaluates small recurrent‑depth language models for multitask clinical text processing, fine‑tuning five models on nine datasets and testing on four unseen datasets. Across 13 tasks, Ouro‑1.4B‑Thinking outperformed Qwen3‑1.7B on 11 datasets, while Ouro‑2.6B‑Thinking matched Qwen3‑8B on most tasks, with the largest gains seen in inference and question answering. The results suggest recurrent‑depth architectures merit further controlled investigation for clinical NLP.
The study compares two fine‑tuning paradigms for detecting hallucinations in large language models: a classifier‑head (“cls”) and a sequence‑to‑sequence (“s2s”) approach. “Cls” consistently outperforms “s2s” because it induces a clear two‑cluster geometry in hidden states, enabling stronger class separation and linear decodability. Analysis of training dynamics reveals a narrow upper‑middle “InfoWindow” where task information emerges; truncating layers above this window removes 31% of the network without loss of accuracy and reduces latency by 1.4×, a strategy that transfers across multiple models and tasks.
The paper introduces Gibson, a virtual environment designed to develop real‑world perception for active agents. Gibson virtualizes real spaces, containing over 1,400 floor plans from 572 buildings, and includes a synthesis mechanism called "Goggles" that allows models trained in simulation to be deployed directly in the real world without additional domain adaptation. The environment also embeds agents within realistic physics and spatial constraints to support embodied learning.
arXiv:2609.32002v2 Announce Type: replace Abstract: The rank $r$ in LoRA is widely treated as a capacity control: a smaller rank is assumed to yield a simpler model that generalizes better. We show t...
arXiv:2607.11578v2 Announce Type: replace-cross Abstract: Deep learning for EEG-based seizure detection faces critical challenges: severe annotation scarcity and extreme class imbalance, where ictal...
arXiv:2609.34540v2 Announce Type: replace Abstract: Ontology Learning (OL) from text has advanced with the emergence of Large Language Models (LLMs), but it remains challenging due to the limited ava...
arXiv:2610.26636v2 Announce Type: replace Abstract: Code generation models are increasingly used to produce interactive artifacts, such as websites and user interfaces. Many properties that determine...
arXiv:1804.08328v2 Announce Type: replace-cross Abstract: Do visual tasks have a relationship, or are they unrelated? For instance, could having surface normals simplify estimating the depth of an im...
Dynamic Cross-Modal Prompt Generation (DRAPE) is a prompt-learning framework designed for Multimodal Continual Instruction Tuning (MCIT). It generates continuous, instance-specific soft prompts by cross-attending textual instructions with visual patch features, rather than selecting from a fixed prompt pool. DRAPE also incorporates null‑space gradient projection to reduce forgetting and uses CLIP‑based prototype routing for task‑label‑free inference, achieving state‑of‑the‑art results on MCIT benchmarks.