Claims Browser

ClusterAttention is the first training-free method that can be successfully applied in the setting of unstructured input and a single forward pass

"To our knowledge, it is the first training-free method that can be successfully applied in the setting of unstructured input and a single forward pass."

For video generation with Wan 2.1-14B T2V, ClusterAttention achieves output closer to dense attention and a larger speedup (1.8x versus 1.4x) compared to SVOO without offline calibration

"For video generation with Wan 2.1-14B T2V arXiv:2503.20314 , ClusterAttention achieves output closer to dense attention and a larger speedup (1.8x versus 1.4x) compared to SVOO arXiv:2603.18636 , a leading method developed specifically for this domain, both run without offline calibration."

Tight clusters can lead to larger errors than random clusters in sparse attention, but when excluded clusters are compensated through their centroids, the error shrinks with tighter clusters

"We also derive an expression for the output error in sparse attention, that explains the counterintuitive experimental finding that tight clusters can lead to larger errors than random clusters. We then derive the error when excluded clusters are compensated through their centroids, and show that this error shrinks with tighter clusters."

Image-to-video generation models have made remarkable progress in subject consistency and temporal coherence, enabling high quality video synthesis in recent years

"In recent years, image-to-video (I2V) generation models have made remarkable progress in subject consistency and temporal coherence, enabling high quality video synthesis."

Advances in I2V models introduce new safety risks that existing studies have largely overlooked, particularly in the temporal dimension

"However, these advances also introduce new safety risks. Existing studies mainly focus on jailbreak attacks involving single frame violations, while largely overlooking the temporal dimension unique to video generation models."

I2V systems have a temporal vulnerability where unsafe semantics can emerge from semantic composition over time rather than from a single frame

"we investigate three attack scenarios and uncover a temporal vulnerability in I2V systems: unsafe semantics may emerge not from a single frame, but from semantic composition over time."

TempJail improves attack success rate over prior state-of-the-art methods by 23.3% under GPT-5.2 evaluation and 22.0% under human evaluation on commercial models

"Experiments on closed-source commercial models, including Kling, Seedance, Veo and PixVerse, show that TempJail improves attack success rate over prior state-of-the-art methods by 23.3\% under GPT-5.2 evaluation and 22.0\% under human evaluation."

Two key challenges in temporal jailbreak attacks are temporal abstraction and semantic camouflage

"We further identify two key challenges in such attacks: temporal abstraction and semantic camouflage."

Representation learning has reached good performances as a pretraining method for downstream tasks and as a first step towards unsupervised speech modeling

"Representation learning has attracted great atten- tion and managed to reach good performances as a pretraining method for downstream tasks or as a first step towards unsu- pervised speech modeling"

Pre-training with limited data exhibits different scaling behavior than web-scale language modeling, where increasing parameters beyond an optimal point causes overfitting rather than improved performance

"Pre-training under limited data requires a different view of scaling than web-scale language modeling. With a fixed data budget but relatively abundant compute, increasing parameter count helps only up to an optimal scale; beyond that point, models overfit and generalization worsens."

Optimal model size depends strongly on both the data budget and the downstream target task

"We study this behavior across 10M-100M word pre-training budgets, two corpora, and multiple downstream evaluations, and find that optimal size depends strongly on both the data budget and the downstream target."

Standard Transformers scale down poorly to limited data settings because embeddings consume too large a fraction of parameters and per-token computation is coupled with representational capacity

"We argue that standard Transformers scale down poorly to this setting, because embeddings consume a large fraction of the parameter budget and per-token computation is tied to representational capacity."

Recursive Transformers with shared blocks and factorized embeddings outperform standard Transformers on limited data budgets of 10M and 100M words

"We train three recursive models and find that they outperform standard Transformers at 10M and 100M words, while remaining competitive with BabyLM Challenge 2025 winners."

Large language model judges are increasingly being used across various evaluation scenarios, making their judgment capabilities valuable intellectual property

"Large language model (LLM) judges are increasingly used across various evaluation scenarios, making their judgment capabilities valuable intellectual property."

Black-box access to LLM judges exposes their capabilities to model extraction attacks

"However, black-box access exposes these capabilities to model extraction attacks."

Existing extraction methods do not specifically target LLM judges and provide limited support for multiple evaluation protocols under restricted query budgets

"Existing extraction methods do not specifically target LLM judges and provide limited support for multiple evaluation protocols under restricted query budgets."

JUDGESTEALER can achieve up to 73.3%, 87.0%, and 71.6% accuracy for pointwise, pairwise, and listwise evaluation respectively when extracting LLM judge capabilities

"Extensive experiments on state-of-the-art LLM-as-a-judge and reward models show that JUDGESTEALER consistently outperforms existing extraction baselines, achieving up to 73.3%, 87.0%, and 71.6% accuracy for pointwise, pairwise, and listwise evaluation, respectively."

JUDGESTEALER demonstrates robustness against representative extraction defenses

"Moreover, JUDGESTEALER demonstrates robustness against representative extraction defenses."