The Optimal Decoder is Tilted Temperature
arXiv:2610.15058v1 Announce Type: new Abstract: Larger sample budgets benefit from higher sampling temperatures in LLMs, but what temperature tuning optimizes and why it works remain poorly understood
Knowledge catalogue
arXiv:2610.15058v1 Announce Type: new Abstract: Larger sample budgets benefit from higher sampling temperatures in LLMs, but what temperature tuning optimizes and why it works remain poorly understood
arXiv:2610.17283v1 Announce Type: new Abstract: The success of autoregressive language models trained with Chain-of-Thought~(CoT) supervision raises a fundamental question: How simple can a next-token
arXiv:2610.16977v1 Announce Type: new Abstract: Visual in-context learning (ICL) enables vision-language models (VLMs) to adapt at test time via demonstration examples, but its effectiveness depends c
arXiv:2610.14265v1 Announce Type: new Abstract: Recent advances in multi-modal large language models have shifted toward the 'thinking with images' paradigm to improve reasoning faithfulness, yet exis
arXiv:2610.15098v1 Announce Type: new Abstract: Text-to-video (T2V) models are highly sensitive to prompt specificity, while iterative prompt optimization typically improves generation quality only th
arXiv:2610.17249v1 Announce Type: cross Abstract: Differentially private (DP) pipelines often combine heterogeneous mechanisms, yet their guarantees are best analyzed using different notions of privac
arXiv:2610.14590v1 Announce Type: cross Abstract: Wheel odometry provides an efficient motion prior for ground robots but is prone to transient violations such as wheel slip or terrain-induced tilt. I
arXiv:2610.14207v1 Announce Type: new Abstract: Under a fixed context budget, which objects and relations should a 3D scene-grounded VLM see first? On ScanNet indoor referring benchmarks (Dai et al.,
arXiv:2610.15807v1 Announce Type: new Abstract: Generating coherent videos grounded in a unified underlying world remains a major challenge, particularly under large viewpoint changes, where existing
arXiv:2602.24111v2 Announce Type: replace Abstract: Vision-language models (VLMs) show promise in drafting radiology reports, yet they frequently suffer from logical inconsistencies, generating diagno
arXiv:2610.16334v1 Announce Type: new Abstract: Robot errors can disrupt human-robot collaboration by reducing user trust and willingness to rely on the robot. Robot apologies can support trust recove
arXiv:2610.16762v1 Announce Type: new Abstract: Keypoint detection plays a central role in geometric vision pipelines. Existing detectors exhibit complementary strengths across different scene structu
arXiv:2610.16857v1 Announce Type: new Abstract: Research on video classification models has witnessed significant growth recently, accompanied by a significant increase in the demand for datasets to t
arXiv:2610.15365v1 Announce Type: cross Abstract: Large language models (LLMs) can produce harmful content while generating a response. A useful safeguard should not only detect this risk during gener
arXiv:2610.14332v1 Announce Type: new Abstract: Accurate sampling from continuous-time generative models can require many model evaluations. Low-step sampling reduces cost, but coarse updates introduc
arXiv:2610.18276v1 Announce Type: new Abstract: Infrared small target detection is often used in different civilian and military applications to detect static or moving targets in scenes like in groun
arXiv:2610.17735v1 Announce Type: new Abstract: Multi-sample reasoning measures candidate coverage, the probability that at least one of K candidates is correct. However, minimizing the average loss o
arXiv:2605.05886v2 Announce Type: replace Abstract: Dense hand contact estimation requires both high-level semantic understanding and fine-grained geometric reasoning of human interaction to accuratel
arXiv:2610.14153v1 Announce Type: new Abstract: Classifier-free guidance (CFG) typically applies a shared scale to all latent tokens, although their velocity predictions evolve differently during samp
arXiv:2409.03945v4 Announce Type: replace Abstract: We present TropNNC, a framework for compressing neural networks with linear and convolutional layers and ReLU-type activations using tropical geomet
arXiv:2610.17530v1 Announce Type: new Abstract: We propose a unified spatial-density signal to score token importance for both inference phases: bipartite merge for prefill and KV eviction for decode,
arXiv:2610.15909v1 Announce Type: cross Abstract: The extension complexity operatorname{xc}(P) of a polytope P is the minimum number of facets of a polytope that projects onto it. Motivated by the com
arXiv:2610.14184v1 Announce Type: new Abstract: Studying model training requires checkpoint trajectories, whose storage cost limits temporal resolution. Although adjacent checkpoints are numerically c
arXiv:2610.15533v1 Announce Type: new Abstract: 3D editing aims to modify an existing 3D asset according to user intent while preserving geometry and appearance outside the intended edit region. Exist
arXiv:2610.16595v1 Announce Type: cross Abstract: Surface-based cortical analysis is useful in many applications as it is a natural way to model the highly folded structure of the cerebral cortex, all
arXiv:2610.15724v1 Announce Type: new Abstract: Denoising Diffusion Probabilistic Models have emerged as powerful generative priors across domains ranging from molecular design to high-resolution imag
arXiv:2610.14394v1 Announce Type: cross Abstract: Training large-scale recommender systems on dynamic industrial platforms, such as Short-Form Video (SFV) services with billions of daily interactions,
arXiv:2610.17699v1 Announce Type: cross Abstract: Single-cell RNA sequencing (scRNA-seq) enables cell-level analysis of biological heterogeneity, but learning effective representations from high-dimen
arXiv:2610.17019v1 Announce Type: cross Abstract: Whisper-style ASR models provide strong zero-shot transcription, but their residual errors are not evenly distributed across datasets, recording condi
arXiv:2507.14777v2 Announce Type: replace Abstract: We investigate foundational questions about the interplay between memorization and learning in large language models (LLMs). When an LLM generates a
arXiv:2610.15564v1 Announce Type: new Abstract: We introduce UNISON, a unified framework that bridges multiple instance learning and recommendation systems by treating both as instances of learning fr
arXiv:2105.10759v4 Announce Type: replace-cross Abstract: We show how a pair of points can uniquely represent a left-infinite sequence obtained from observations of an underlying dynamical system thro
arXiv:2601.17274v2 Announce Type: replace Abstract: We develop unrolled neural networks to solve constrained optimization problems, offering accelerated, learnable counterparts to dual ascent (DA) alg
arXiv:2511.13714v2 Announce Type: replace Abstract: The Segment Anything Model (SAM) family has become a widely adopted vision foundation model, but its ability to control segmentation granularity rem
arXiv:2610.18291v1 Announce Type: new Abstract: Since the rise of supervised deep learning modelet dets, researchers has attempt to solve infrared small targection using various models like convention
arXiv:2610.16826v1 Announce Type: cross Abstract: Deepfake speech detectors are typically evaluated as binary classifiers on fixed benchmarks, but deployed systems face evolving attacks, codecs, speak
arXiv:2610.16578v1 Announce Type: cross Abstract: Modern deep classifiers trained by gradient descent need many passes over the data to become accurate, and even then are not always well calibrated; p
arXiv:2610.16204v1 Announce Type: new Abstract: Autoencoders are the core component of latent video generative models, compressing high-dimensional videos into compact latent spaces for downstream gen
arXiv:2610.18615v1 Announce Type: new Abstract: Visual agents must turn video observations into timely decisions to track progress and choose what to do next. Building on the emerging System One parad
arXiv:2610.15728v1 Announce Type: new Abstract: Visible watermarks convey document ownership, confidentiality, and status, but can also carry adversarial signals. We introduce Auto-Regional Adversaria
arXiv:2610.16567v1 Announce Type: new Abstract: Vision-Language-Action models (VLAs) have achieved high success rates on familiar manipulation tasks, yet it remains unclear whether such scores reflect
arXiv:2610.15092v1 Announce Type: new Abstract: Quantization plays a crucial role in efficiently deploying deep learning models on resource-constrained devices. Post-training quantization does not req
arXiv:2610.18147v1 Announce Type: new Abstract: We present a real-time framework for expressive legged locomotion that enables independent control over velocity and style. Unlike existing data-driven
arXiv:2603.27158v3 Announce Type: replace Abstract: While highly accelerated non-Cartesian acquisition protocols significantly reduce scan time, they often entail long reconstruction delays. Deep lear
arXiv:2610.15850v1 Announce Type: new Abstract: Sparse autoencoder (SAE) features are typically discovered from how their activations co-vary with behaviour, yet they are often used through interventi
arXiv:2610.16641v1 Announce Type: new Abstract: Cross-task generalization remains difficult for world action models (WAMs), even when their predicted futures depict plausible manipulation. This gap ca
arXiv:2610.17844v1 Announce Type: new Abstract: Spatial understanding is a necessary step in moving multimodal large language models (MLLMs) toward reasoning about the 3D physical world. Their visual
arXiv:2610.15035v1 Announce Type: new Abstract: Contrastive learning has become the standard paradigm for multimodal representation learning, aligning semantically matched pairs while separating misma
arXiv:2610.18155v1 Announce Type: new Abstract: Standard identifiability theory assumes one-dimensional latent factors, but learned representations can encode latent information across multidimensiona
arXiv:2610.16020v1 Announce Type: new Abstract: Layer-wise post-training quantization methods let each layer compensate for the quantization error inherited from earlier layers. We ask under what assu
arXiv:2605.06204v2 Announce Type: replace-cross Abstract: Trimming suspicious calibration points is a common response to contamination in conformal prediction. Its effect on clean-target coverage, how
arXiv:2610.15139v1 Announce Type: new Abstract: Linear-attention and state-space LLMs carry long context in a fixed-size recurrent state, and quantizing that state is the natural way to cut its cost d
arXiv:2610.16220v1 Announce Type: new Abstract: Participant-disjoint cross-validation prevents direct participant overlap, but it may still produce optimistic estimates when acquisition site and diagn
arXiv:2610.14090v1 Announce Type: new Abstract: Exterior Gaussian-splat twins for real-estate listings are captured to a fixed recipe: a consumer drone orbits a single property in several overlapping
arXiv:2610.16801v1 Announce Type: cross Abstract: Tabular foundation models (TFMs) achieve state-of-the-art performance on IID benchmarks, but how do they respond to repeated measurements at the same
arXiv:2610.17791v1 Announce Type: new Abstract: Understanding facial behaviour in video requires answering three fundamental questions: when a key facial expression occurs, where the model attends on
arXiv:2610.17285v1 Announce Type: new Abstract: Modern neural-network training uses mixed recipes, assigning AdamW-like or sign-based updates to embedding and output layers and spectral updates to hid
arXiv:2610.15299v1 Announce Type: new Abstract: Misclassification is a critical problem in many machine learning applications. Since even the classifier models with high accuracy (e.g., > 95%) still i
arXiv:2610.16337v1 Announce Type: new Abstract: Rotary Positional Embedding (RoPE) is widely used to encode relative positions in Transformers. In multi-view settings, however, patch coordinates are d
I’m prioritizing my path and was wondering if an MS is a soft requirement for some of these Research Engineer, Applied Scientist, Post-Training roles. I only have a BS but I work full time as a Data S