Huang R, Xiong Y, Cheng YY et al.
Parameter-efficient fine-tuning (PEFT) methods adapt large models by introducing a small number of trainable parameters.Parameter-efficient fine-tuning (PEFT) methods adapt large models by introducing a small number of trainable parameters. However, we find that these parameters still contain redundancy and can be further compressed while…
Zou J, Li X, Yue Z et al.
Discrete visual tokens represent vision in a discrete form analogous to language, offering a native interface for multimodal modeling under a shared next-token prediction…Discrete visual tokens represent vision in a discrete form analogous to language, offering a native interface for multimodal modeling under a shared next-token prediction objective.
Shilov I, Yang X, Panfilov A et al.
We introduce Heads of Security (HoS), a prompt injection defense that enforces instruction-data separation via a small set of learned per-head attention gates.We introduce Heads of Security (HoS), a prompt injection defense that enforces instruction-data separation via a small set of learned per-head attention gates.
Chuang I, Shah H, Chilamkur V et al.
Dexterous manipulation with multi-fingered hands is the next frontier of robotics, yet learned hand skills remain limited to simple grasps or basic repeated motions, such…Dexterous manipulation with multi-fingered hands is the next frontier of robotics, yet learned hand skills remain limited to simple grasps or basic repeated motions, such as rotating an object.
Bun M, Juvekar M
The sign rank of a ± 1-valued matrix A is the minimum rank of a real matrix that agrees with A entrywise in sign.The sign rank of a ± 1-valued matrix A is the minimum rank of a real matrix that agrees with A entrywise in sign. For every sufficiently large d, we give an explicit matrix A with 2^(Θ(d^2)) rows and columns and sign rank Θ(d)…
Aggarwal P, Welleck S
AI agents have made substantial progress in domains where they have a non-trivial amount of time to deliberate between actions, such as mathematics, software engineering…AI agents have made substantial progress in domains where they have a non-trivial amount of time to deliberate between actions, such as mathematics, software engineering, and performing tasks on desktop applications.
Lin J, Zhang Y, Zhang H et al.
Predictive maintenance is vital for commercial aviation, yet progress remains hindered by the lack of public, large-scale real-world datasets.Predictive maintenance is vital for commercial aviation, yet progress remains hindered by the lack of public, large-scale real-world datasets.
Ding P
Label-free reinforcement learning with verifiable rewards (RLVR) rewards a language model for agreeing with its own majority vote.Label-free reinforcement learning with verifiable rewards (RLVR) rewards a language model for agreeing with its own majority vote. It improves reasoning at first but eventually collapses: the self-assessed reward goes to one while…
Ou D, Lopez A, Missinne Q et al.
Event cameras offer low latency, high dynamic range, and milliwatt-level power consumption, making them ideal for agile robots such as tiny drones navigating through…Event cameras offer low latency, high dynamic range, and milliwatt-level power consumption, making them ideal for agile robots such as tiny drones navigating through cluttered environments.
Fink E
Self-attention mechanisms have established the Transformer as the dominant backbone across natural language processing, source code intelligence, and automated theorem…Self-attention mechanisms have established the Transformer as the dominant backbone across natural language processing, source code intelligence, and automated theorem proving.
Reddy S, Abbas A, Suri A et al.
Fine-tuning Large Language Models (LLMs) on instruction data can quietly erode their safety, even when no training example looks harmful.Fine-tuning Large Language Models (LLMs) on instruction data can quietly erode their safety, even when no training example looks harmful. Because this safety-degrading data looks benign, standard content filters struggle to catch…
Bekos P, Tsouvalas B, Nikiforakis N et al.
AI chatbot platforms have become one of the most popular web interfaces users interact with online, routinely receiving personal and intimate information shared under an…AI chatbot platforms have become one of the most popular web interfaces users interact with online, routinely receiving personal and intimate information shared under an expectation of privacy.
Sprengers J, Bognar M, Daniel LA et al.
Data memory-dependent prefetchers (DMPs) are a microarchitectural optimization recently introduced in mainstream desktop and server CPUs.Data memory-dependent prefetchers (DMPs) are a microarchitectural optimization recently introduced in mainstream desktop and server CPUs. Although DMPs offer performance improvements, they have been shown to enable side-channel…
G\"urb\"uzbalaban M, Rafi MH, Aybat NS et al.
With exact, delay-free gradients, Nesterov's accelerated gradient (NAG) achieves optimal √κ dependence for L-smooth, μ-strongly convex objectives, where κ=L/μ; heavy-ball…With exact, delay-free gradients, Nesterov's accelerated gradient (NAG) achieves optimal √κ dependence for L-smooth, μ-strongly convex objectives, where κ=L/μ; heavy-ball (HB) achieves the same scaling on strongly convex…
Baniecki H, Biecek P, Pawelczyk M
Language model performance depends strongly on the mixture of domains in the pretraining data corpus. Popular data mixing methods select these domain weights by training…Language model performance depends strongly on the mixture of domains in the pretraining data corpus. Popular data mixing methods select these domain weights by training many small proxy models on random mixes and then fitting a…
Kim G, Hwang SJ
Mixture-of-Experts (MoE) models are bottlenecked by loading routed-expert weights from memory during interactive decoding at small batch sizes.Mixture-of-Experts (MoE) models are bottlenecked by loading routed-expert weights from memory during interactive decoding at small batch sizes.
Vera HS, Dua S, Andonov J et al.
We present EmbeddingGemma 2, an open multimodal embedding model which natively represents text, images, audio, video, and combinations thereof in a unified space.We present EmbeddingGemma 2, an open multimodal embedding model which natively represents text, images, audio, video, and combinations thereof in a unified space.
Ying W, Liu D, Sigal L et al.
Existing evaluations of text-conditioned 3D indoor scene generation largely emphasize semantic fidelity and physical validity at rest, leaving interaction-time physical…Existing evaluations of text-conditioned 3D indoor scene generation largely emphasize semantic fidelity and physical validity at rest, leaving interaction-time physical reliability insufficiently assessed.
Chang H, Elsahar H, Sander T et al.
Proprietary language models and their reasoning traces are vulnerable to API-based distillation, where an adversary trains a student model on the teacher model's outputs.Proprietary language models and their reasoning traces are vulnerable to API-based distillation, where an adversary trains a student model on the teacher model's outputs.
Dongmo V, Ninyim AMF
Multi-relational graph detectors learn from clean data how much to trust each relation, but an adversary pays a different price to manipulate each relation and each…Multi-relational graph detectors learn from clean data how much to trust each relation, but an adversary pays a different price to manipulate each relation and each action; taking such costs as given, we ask when using a relation…
Guo J, Miao X
The fastest machine-learning kernels are compositions of components, such as loaders, matrix-multiply mainloops, epilogues and communication stages, that run concurrently…The fastest machine-learning kernels are compositions of components, such as loaders, matrix-multiply mainloops, epilogues and communication stages, that run concurrently on one GPU streaming multiprocessor and share its on-chip…
Wang S, Zhao W, Lu J et al.
In finance, the answer that matters is usually the current one: prices, rates, and disclosures hold only as of a given day, so retrieving up-to-date facts is a core…In finance, the answer that matters is usually the current one: prices, rates, and disclosures hold only as of a given day, so retrieving up-to-date facts is a core capability of a financial agent.
Curci G, Simonazzi S, Molinari A et al.
Direct Preference Optimization (DPO) typically relies on preference pairs generated from diverse model pools and ranked by an external LLM judge.Direct Preference Optimization (DPO) typically relies on preference pairs generated from diverse model pools and ranked by an external LLM judge.
Zhu W, Zuo M, Sun H
Audit-based certification bounds the violation rate of an LLM agent by auditing a random fraction ε of actions and reweighting the audited labels; its validity is meant…Audit-based certification bounds the violation rate of an LLM agent by auditing a random fraction ε of actions and reweighting the audited labels; its validity is meant to rest only on the defender's audit coin.
Cao S, Mao Z, Li D et al.
The rapid progress of software engineering agents drives interest in applying them to open-ended problems such as GPU kernel optimization.The rapid progress of software engineering agents drives interest in applying them to open-ended problems such as GPU kernel optimization. However, we find that existing coding agents converge to local optima on real-world…
Yuan X, Yao B, Yang L et al.
Leading GUI grounders score 93 to 96% on ScreenSpot-v2, and reported gaps between new and prior models are about a point.Leading GUI grounders score 93 to 96% on ScreenSpot-v2, and reported gaps between new and prior models are about a point. At this level the useful question is what the remaining errors are.
Yang Z, Kumar R, Kroer C
We consider T repeated second-price auctions across n pacing bidders with i.i.d. item arrivals, and show that it is computationally impossible for all bidders to achieve…We consider T repeated second-price auctions across n pacing bidders with i.i.d. item arrivals, and show that it is computationally impossible for all bidders to achieve no-regret simultaneously while satisfying their respective…
Ziv L, Sharif M
While few-step distilled text-to-image models have gained traction due to their fast generation times, the speed-up implications on their unsafe generation rates remain…While few-step distilled text-to-image models have gained traction due to their fast generation times, the speed-up implications on their unsafe generation rates remain unknown.
Deng Y, Ruan J, Huang Y et al.
Coding agents are beginning to develop the production-grade AI infrastructure that trains and serves large models, yet existing benchmarks cover only single kernels or…Coding agents are beginning to develop the production-grade AI infrastructure that trains and serves large models, yet existing benchmarks cover only single kernels or single-GPU tasks, and grade with tests and prompts that…
Liu YP, Naidu K, Nguyen HA et al.
We prove the first unconditional lower bound tradeoff between the number of passes and the approximation factor of maximum matchings in the semi-streaming model.We prove the first unconditional lower bound tradeoff between the number of passes and the approximation factor of maximum matchings in the semi-streaming model.
Liu B, Ye J, Yuan X et al.
Recently, LLM-based autonomous agents have shown strong capabilities in iteratively generating, evaluating, and refining executable solutions.Recently, LLM-based autonomous agents have shown strong capabilities in iteratively generating, evaluating, and refining executable solutions.
Pan J, Lyu Y, Hou AB et al.
LLM agents increasingly interact with other agents on behalf of users, yet how users in the loop affect their agents' behavior–for better or worse–remains poorly…LLM agents increasingly interact with other agents on behalf of users, yet how users in the loop affect their agents' behavior–for better or worse–remains poorly understood.
Chen Y, Peng D, Zhu Z et al.
A small number of GPU kernels often determines the efficiency of large language model training and inference.A small number of GPU kernels often determines the efficiency of large language model training and inference.
Hickok T
Low-cost arms such as the SO-101 have become the most common entry point to robot learning, yet no released policy performs spatially constrained, language-conditioned…Low-cost arms such as the SO-101 have become the most common entry point to robot learning, yet no released policy performs spatially constrained, language-conditioned tasks on them out of the box, and existing SO-101 benchmarks…
Zhang B, Chu W, Li Y et al.
We present Scalable Pixel-anchored End-to-end Diffusion (SpeeDiff), a latent diffusion method that jointly trains the VAE and the diffusion model from scratch.We present Scalable Pixel-anchored End-to-end Diffusion (SpeeDiff), a latent diffusion method that jointly trains the VAE and the diffusion model from scratch.
Brom PC
Typed decision models return a probability distribution over options supplied by the caller, and software tends to act on that number as if it were a guarantee.Typed decision models return a probability distribution over options supplied by the caller, and software tends to act on that number as if it were a guarantee.
Van MP, Tran C
A retrieval-augmented language model answers from two sources, the retrieved context and its own knowledge, and the two disagree whenever the context is incomplete or…A retrieval-augmented language model answers from two sources, the retrieved context and its own knowledge, and the two disagree whenever the context is incomplete or wrong.
Paina M, M\"uller MN, M\"undler-Sasahara N et al.
Proactive bug discovery is an important software-engineering problem. Even though coding agents are already widely deployed in this domain, their capabilities remain…Proactive bug discovery is an important software-engineering problem. Even though coding agents are already widely deployed in this domain, their capabilities remain understudied.
Kovalenko A, Redchych A
Small open language models can call tools, but they do not reason about what a call does: they delete without asking, repeat writes, re-read state they already hold, and…Small open language models can call tools, but they do not reason about what a call does: they delete without asking, repeat writes, re-read state they already hold, and issue one call per turn where a short program would do.
Srinivasan S, Li L, Menzies T et al.
Early stopping promises to reduce compute costs for autonomous coding agents by terminating unpromising runs early.Early stopping promises to reduce compute costs for autonomous coding agents by terminating unpromising runs early. Deciding when to abandon a run requires separating baseline issue difficulty from the trajectory of an individual…