What Should a Streaming Video Model Remember?
arXiv:2606.16353v2 Announce Type: replace Abstract: Streaming video understanding models must answer queries at any moment during an ongoing stream, using only what they have observed so far and under
Knowledge catalogue
arXiv:2606.16353v2 Announce Type: replace Abstract: Streaming video understanding models must answer queries at any moment during an ongoing stream, using only what they have observed so far and under
arXiv:2609.32991v1 Announce Type: new Abstract: What should data teach a language model at a particular point in training? A circuit view reveals three distinct bottlenecks: forming a computation, mak
arXiv:2609.32485v1 Announce Type: new Abstract: Federated low-rank adaptation (LoRA) requires identifying an update structure that is shared across heterogeneous clients. Prior work reports strong sim
arXiv:2609.32452v1 Announce Type: cross Abstract: Reflective prompt optimization revises instructions using examples of a model's behavior, but which evidence the reflector should receive remains uncl
arXiv:2609.32124v1 Announce Type: new Abstract: When adapting pre-trained models through fine-tuning, freezing weights alone might not preserve performance, as updates elsewhere can change the inputs
arXiv:2609.31760v1 Announce Type: cross Abstract: Can a robot improve itself the way coding agents now improve software? We built an agentic system to find out. It watches a robot fail, works out whic
arXiv:2609.33375v1 Announce Type: cross Abstract: The transition from vocoder-based to neural-codec speech synthesis makes generalization more difficult for speech deepfake detectors, particularly tho
arXiv:2609.34732v1 Announce Type: new Abstract: Representation alignment speeds up diffusion transformer training by pulling an intermediate block of the model (student) toward features of a frozen pr
arXiv:2609.32670v1 Announce Type: cross Abstract: When a language model explains an answer it has already given, does it reuse the computation that produced the answer or reconstruct a story from the
arXiv:2609.33189v1 Announce Type: cross Abstract: When the human reference scores zero on a metric, the released GTRS-Dense label generator for NAVSIM marks every candidate trajectory in the scene as
arXiv:2609.32677v1 Announce Type: cross Abstract: Self-improving agents increasingly rely on proxy verifiers to choose policy updates, yet deployment can change the world in which those updates are ev
arXiv:2609.34650v1 Announce Type: new Abstract: Some attention heads learn similar patterns across inputs. Reusing these patterns could reduce training cost by avoiding repeated query-key score comput
arXiv:2609.32267v1 Announce Type: new Abstract: Releasing a model update requires certifying that its current-population risk stays below a threshold. Trusted labels are expensive, while a cheap evalu
arXiv:2606.18383v2 Announce Type: replace Abstract: Sparse autoencoders (SAEs) are increasingly used to study internal representations in language models (LMs), but it remains unclear when an SAE-base
arXiv:2609.34771v1 Announce Type: cross Abstract: Reliable AI safeguards require both control mechanisms that reduce unsafe behavior and monitoring mechanisms that detect safety risks during model int
arXiv:2609.33220v1 Announce Type: new Abstract: Outcome-based reinforcement learning can produce models with similar task performance but very different ways of communicating about their mistakes. We
arXiv:2504.12860v2 Announce Type: replace-cross Abstract: We study the effectiveness of randomizing split-directions in random forests. Prior literature has shown that, randomization can reduce varian
arXiv:2609.34480v1 Announce Type: new Abstract: Reliable visual question answering requires correct answers when evidence is sufficient and abstention when it is not. We introduce a benchmark that con
arXiv:2609.33169v1 Announce Type: new Abstract: Policies with memory can learn along two backward paths: through the physical states their actions produce and through the representations they store. T
arXiv:2609.32488v1 Announce Type: cross Abstract: Dense retrieval powers retrieval-augmented generation, semantic search, and question answering, yet the theoretical basis for choosing between shared
arXiv:2606.16694v2 Announce Type: replace Abstract: Transformers perform computations through many successive attention and feedforward blocks, allowing them to learn complex correlations between a la
arXiv:2609.31639v1 Announce Type: new Abstract: Diagnosing rolling-element bearing faults from vibration is a canonical physical-sensing task and a widely used benchmark for domain adaptation under op
arXiv:2609.33359v1 Announce Type: new Abstract: Geometric view synthesis can expand a single wine-label photograph into a training set, but its value with pretrained image encoders is unclear. We stud
arXiv:2609.34460v1 Announce Type: cross Abstract: Does structured mathematical knowledge help LLMs prove theorems in Lean 4? If so, for which models, and does the answer vary by problem? Formal librar
arXiv:2609.32301v1 Announce Type: new Abstract: Active feature acquisition (AFA) sequentially selects informative features under budget constraints. However, existing policies rarely distinguish wheth
arXiv:2609.32489v1 Announce Type: cross Abstract: In tri-modal visual question answering (VQA), auxiliary text is commonly used to complement visual and textual inputs, yet its reliability is often un
arXiv:2609.34091v1 Announce Type: new Abstract: Sparse autoencoders (SAEs) decompose dense model activations into discrete latents, making individual features easy to interpret--and easy to misinterpr
arXiv:2609.31678v1 Announce Type: new Abstract: KV cache compression is widely used for long context LLM inference under memory constraints, while deployed systems typically score refusals after gener
arXiv:2609.34012v1 Announce Type: new Abstract: Neural PDE surrogates increasingly incorporate structural priors, yet it is often unclear whether their gains arise from physics-specific information or
arXiv:2609.32898v1 Announce Type: new Abstract: Pretrained tabular foundation models process every dataset at a fixed depth, with inference costs growing with dataset size. To address this, we present
arXiv:2609.34489v1 Announce Type: new Abstract: Data pruning reduces the training cost of knowledge distillation (KD). However, the preferred teacher capacity changes with the data budget: smaller tea
arXiv:2609.34602v1 Announce Type: new Abstract: Frozen Earth-observation embeddings are judged almost entirely by spatially blocked cross-validation inside one study region. We show that this number d
arXiv:2609.33668v1 Announce Type: new Abstract: Pseudo-labeling has become a cornerstone of learning from unlabeled data in semantic segmentation. Yet its effectiveness drops sharply in real-world sce
arXiv:2609.33312v1 Announce Type: cross Abstract: Moving ML-mediated decision making onto privacy-preserving clients decentralises the economic decision along with the inference. Shared budget constra
arXiv:2609.31733v1 Announce Type: new Abstract: Retrieval-augmented generation is an attractive way to improve chest X-ray reporting, because reports from similar prior studies supply clinical context
arXiv:2609.32083v1 Announce Type: new Abstract: Deploying AI systems requires deciding when to delegate tasks and when humans should intervene to monitor and mitigate risk induced by AI operations. Th
arXiv:2609.35314v1 Announce Type: new Abstract: Neural residuals can improve satellite evapotranspiration (ET) estimates, but selectors must predict when a correction helps and reject unsupported inpu
arXiv:2603.13688v2 Announce Type: replace-cross Abstract: AI systems increasingly assist decision making by producing cheap factor-level assessments of complex inputs, but these assessments can be bia
arXiv:2609.35416v1 Announce Type: new Abstract: Continuous video counting requires distinguishing new observations from new objects or completed events. We introduce StaMina (State Maintenance), which
arXiv:2609.34849v1 Announce Type: new Abstract: Reinforcement learning with verifiable rewards provides a sparse post-training signal: a single binary outcome evaluates the entire rollout, and every t
arXiv:2609.34861v1 Announce Type: new Abstract: Visual token pruning has been widely studied as a practical approach to reducing the computational cost of large vision-language models. However, it str
arXiv:2609.32332v1 Announce Type: new Abstract: Rehearsal-free continual learning with parameter-efficient adapters can be cast as a sequence of task-vector write-in operations: for each new task, a l
arXiv:2609.34440v1 Announce Type: new Abstract: Driving benchmark scores are increasingly used not only for evaluation but also as optimization targets. This raises a fundamental question: do score ga
arXiv:2606.18820v2 Announce Type: replace Abstract: Sequential decisions often become better informed as action opportunities expire. We introduce Maturing Markov Decision Processes (MMDPs), a structu
arXiv:2609.33334v1 Announce Type: new Abstract: Training-free KV-cache compression methods such as SnapKV, H2O, and PyramidKV evict tokens at the end of prefill, aiming to preserve the attention mass
arXiv:2609.34781v1 Announce Type: new Abstract: Vision-language models (VLMs) can read text in natural scenes, but their predictions may be influenced by the surrounding context. When the printed text
arXiv:2609.34300v1 Announce Type: new Abstract: World models offer a powerful substrate for safety reasoning in high-dimensional robotic systems, but they are also fallible: their predictions can be b
arXiv:2609.33889v1 Announce Type: new Abstract: At each step, decoding one sequence with a large language model rereads the projection weights, whose traffic is fixed, and the key-value (KV) cache, wh
arXiv:2609.32230v1 Announce Type: new Abstract: Surrogate models are commonly evaluated by how often they agree with their teacher model over an evaluation set. Local variation in this agreement is we
arXiv:2609.31722v1 Announce Type: new Abstract: Fixed image distortions do not cover an attacker that learns from paired clean and watermarked images. We study this paired-training threat with single-
arXiv:2609.34554v1 Announce Type: new Abstract: Memory is essential for long-horizon, partially observed robotic manipulation: a robot must remember which object was placed in a drawer, whose cup it m
arXiv:2609.33180v1 Announce Type: cross Abstract: As recursive self-improvement (RSI) rapidly advances, reliable evaluation becomes critical for guiding adaptive search. RSI typically relies on finite
arXiv:2609.35630v1 Announce Type: cross Abstract: Massive activation features (MAs) in Transformers are extreme-value residual-stream features that persist across layers despite the model's ability to
arXiv:2609.34078v1 Announce Type: new Abstract: Domain adaptation is crucial for addressing distributional shifts that degrade model performance across domains. While most existing research has center
arXiv:2609.34065v1 Announce Type: cross Abstract: Names are personal identifiers, but they also carry social meaning and are widely used to evaluate how language models treat different people. Such ev
arXiv:2609.35486v1 Announce Type: cross Abstract: High instance-level accuracy can mask inconsistencies in spatial reasoning when objects exchange positions or their roles are reversed in the query. T
arXiv:2505.15064v5 Announce Type: replace Abstract: We ask when adding hidden layers improves generalization, in a model that keeps the layers fixed and varies only their number. A hidden layer is a s
arXiv:2509.23219v2 Announce Type: replace Abstract: Technical-domain benchmarks constructed from arXiv papers can overlap the same public text used in LLM pretraining. Auditing this risk at training-c
arXiv:2609.34826v1 Announce Type: new Abstract: Humans often solve spatial problems by mentally simulating visual transformations. In contrast, conventional vision-language models (VLMs) reason primar
arXiv:2609.34138v1 Announce Type: cross Abstract: With the advent of word representations, word similarity tasks are becoming increasing popular as an evaluation metric for the quality of the represen