We thereby identified the offset effect, a phenomenon characterized by two key findings: (1) verbatim memorization is most strongly triggered by short prefixes drawn from the beginning of the context window, with memorization decreasing counterintuitively as prefix length increases; and (2) a sharp decline in verbatim recall when prefix begins offset from the initial tokens of the context window. We attribute this to positional fragility: models rely disproportionately on the earliest tokens in their context window as retrieval anchors, making them sensitive to even slight shifts. We further observe that when the model fails to retrieve memorized content, it often produces degenerated text. Leveraging these findings, we show that shifting sensitive data deeper into the context window suppresses both extractable memorization and degeneration. Our results suggest that positional offset is a critical and previously overlooked axis for evaluating memorization risks, since prior work implicitly assumed uniformity by probing only from the beginning of documents or training sequences.

Large language models (LLMs) have achieved remarkable progress in natural language generation, yet they continue to display puzzling behaviors--such as repetition and incoherence--even when exhibiting low perplexity. This highlights a key limitation of conventional evaluation metrics, which emphasize local prediction accuracy while overlooking long-range structural complexity. We introduce correlation dimension, a fractal-geometric measure of self-similarity, to quantify the epistemological complexity of text as perceived by a language model.

Test-time adaptation (TTA) aims to adapt a source model to a target domain using only test data. Existing methods predominantly rely on unsupervised entropy minimization or its variants, which suffer from degeneration, leading to trivial solutions with low-entropy but inaccurate predictions. In this work, we identify entropy-deceptive (ED) samples, instances where the model makes highly confident yet incorrect predictions, as the underlying cause of degeneration. Further, we reveal that the gradients of entropy minimization in TTA have an intrinsic lowdimensional structure, driven primarily by entropy-truthful (ET) samples whose gradients are highly correlated. In contrast, ED samples have scattered, less correlated gradients. Leveraging this observation, we show that the detrimental impact of ED samples can be suppressed by constraining model updates within the principal subspace of backward gradients. Building on this insight, we propose LCoTTA, a lifelong continual TTA method that tracks the principal subspace of gradients online and utilizes their projections onto this subspace for adaptation. Further, we provide theoretical analysis to show that the proposed subspace-based method can enhance the robustness against detrimental ED samples. Extensive experiments demonstrate that LCoTTA effectively overcomes degeneration and significantly outperforms existing methods in long-term continual adaptation scenarios.

The world's longest-living vertebrates maintain their vision for centuries. Breakthroughs, discoveries, and DIY tips sent every weekday. The Greenland shark () is well-known for its impressive lifespan. Marine biologists believe the world's longest-living vertebrate often reaches over 400 years old, and possibly lives even longer. But while the shark isn't known for its vision, a lot could be learned from the deep-sea predator's eyes. According to new research recently published in the journal, the Greenland shark retained its visual organs throughout millions of years of evolution for a reason.

This paper proposes a new enhanced model architecture to perform classification of lumbar spine degeneration with DICOM images while using a hybrid approach, integrating EfficientNet and VGG19 together with custom-designed components. The proposed model is differentiated from traditional transfer learning methods as it incorporates a Pseudo-Newton Boosting layer along with a Sparsity-Induced Feature Reduction Layer that forms a multi-tiered framework, further improving feature selection and representation. The Pseudo-Newton Boosting layer makes smart variations of feature weights, with more detailed anatomical features, which are mostly left out in a transfer learning setup. In addition, the Sparsity-Induced Layer removes redundancy for learned features, producing lean yet robust representations for pathology in the lumbar spine. This architecture is novel as it overcomes the constraints in the traditional transfer learning approach, especially in the high-dimensional context of medical images, and achieves a significant performance boost, reaching a precision of 0.9, recall of 0.861, F1 score of 0.88, loss of 0.18, and an accuracy of 88.1%, compared to the baseline model, EfficientNet. This work will present the architectures, preprocessing pipeline, and experimental results. The results contribute to the development of automated diagnostic tools for medical images.

Large language models have significantly advanced Multilingual Machine Translation (MMT), yet the broad language coverage, consistent translation quality, and English-centric bias remain open challenges. To address these challenges, we introduce \textbf{LMT}, a suite of \textbf{L}arge-scale \textbf{M}ultilingual \textbf{T}ranslation models centered on both Chinese and English, covering 60 languages and 234 translation directions. During development, we identify a previously overlooked phenomenon of \textbf{directional degeneration}, where symmetric multi-way fine-tuning data overemphasize reverse directions (X $\to$ En/Zh), leading to excessive many-to-one mappings and degraded translation quality. We propose \textbf{Strategic Downsampling}, a simple yet effective method to mitigate this degeneration. In addition, we design \textbf{Parallel Multilingual Prompting (PMP)}, which leverages typologically related auxiliary languages to enhance cross-lingual transfer. Through rigorous data curation and refined adaptation strategies, LMT achieves SOTA performance among models of comparable language coverage, with our 4B model (LMT-60-4B) surpassing the much larger Aya-101-13B and NLLB-54B models by a substantial margin. We release LMT in four sizes (0.6B/1.7B/4B/8B) to catalyze future research and provide strong baselines for inclusive, scalable, and high-quality MMT \footnote{\href{https://github.com/NiuTrans/LMT}{https://github.com/NiuTrans/LMT}}.

Large language models (LLMs) have achieved remarkable progress in natural language generation, yet they continue to display puzzling behaviors -- such as repetition and incoherence -- even when exhibiting low perplexity. This highlights a key limitation of conventional evaluation metrics, which emphasize local prediction accuracy while overlooking long-range structural complexity. We introduce correlation dimension, a fractal-geometric measure of self-similarity, to quantify the epistemological complexity of text as perceived by a language model. This measure captures the hierarchical recurrence structure of language, bridging local and global properties in a unified framework. Through extensive experiments, we show that correlation dimension (1) reveals three distinct phases during pretraining, (2) reflects context-dependent complexity, (3) indicates a model's tendency toward hallucination, and (4) reliably detects multiple forms of degeneration in generated text. The method is computationally efficient, robust to model quantization (down to 4-bit precision), broadly applicable across autoregressive architectures (e.g., Transformer and Mamba), and provides fresh insight into the generative dynamics of LLMs.

Large Language Models (LLMs) often generate repetitive and monotonous outputs, especially in tasks like story generation, due to limited creative diversity when given the same input prompt. To address this challenge, we propose a novel decoding strategy, Avoidance Decoding, that modifies token logits by penalizing similarity to previously generated outputs, thereby encouraging more diverse multi-branch stories. This penalty adaptively balances two similarity measures: (1) Concept-level Similarity Penalty, which is prioritized in early stages to diversify initial story concepts, and (2) Narrative-level Similarity Penalty, which is increasingly emphasized later to ensure natural yet diverse plot development. Notably, our method achieves up to 2.6 times higher output diversity and reduces repetition by an average of 30% compared to strong baselines, while effectively mitigating text degeneration. Furthermore, we reveal that our method activates a broader range of neurons, demonstrating that it leverages the model's intrinsic creativity.