cập nhật 7 phút trước

⏱ Độ trễ thật của tab này: trung bình 6.8 giờ sau khi CVE được công bố (n = 5267, sớm nhất 0.1 giờ). Số âm nghĩa là ta thấy trước ngày công bố.

26
Modeling LLM Reasoning Dynamics with Structural Representations
📄 Paper arXiv cs.AI 17 giờ trước

arXiv:2611.23749v1 Announce Type: new Abstract: Language-model reasoning alternates between sustained activities and local operations. We study how this temporal structure predicts continuation and the computation needed to analyze it. T/E/S combines text, local events, and span…

🅲 hạng C
25
CADMAGEDDON: High-Fidelity CAD Reconstruction with Local Inference
📄 Paper HF Daily Papers 3 ngày trước

CAD practitioners often work with proprietary parts that they are unwilling to share with external services. For individuals and small companies, bringing this work in-house also requires computational resources they can afford. We introduce CADMAGEDDON, an open-source agent that…

🅲 hạng C
25
From Token Changes to Task Failures: Bit Flips in a Shared KV-Cache Block
📄 Paper arXiv cs.CR 1 ngày trước

arXiv:2611.22243v1 Announce Type: new Abstract: In LLM serving systems with prefix caching, each KV block of a shared system prompt is held in GPU memory as a single physical copy, which every request with that prompt reads. Our earlier work measured change and showed that one b…

🅲 hạng C
24
Termination Lag: Coding Agents Solve Early but Stop Late
📄 Paper arXiv cs.AI 6 ngày trước

arXiv:2611.14960v1 Announce Type: new Abstract: Coding agents can solve a task well before they stop working on it. We characterize \emph{termination lag}--the number of steps between an agent's first correct patch and the end of its run--across five models, two agent frameworks…

🅲 hạng C
24
Attacker Diversity Reduction in Adaptive LLM Red Teaming
📄 Paper arXiv cs.CR 2 ngày trước

arXiv:2611.19680v1 Announce Type: new Abstract: Reinforcement learning can make automated red-teaming more successful while narrowing the attacks it produces. Co-training an attacker and a defender with group-relative policy optimization (GRPO) has been reported to reduce attack…

🅲 hạng C
23
Calpha: Toward Native Biomolecular Intelligence in a General-Purpose Language Model
📄 Paper HF Daily Papers 4 ngày trước

Biomolecular foundation models predict the structures of proteins and their complexes accurately, and large language models (LLMs) reason in natural language, but the two remain separate: the abilities of the former are not part of a general-purpose LLM, and the reasoning of the …

🅲 hạng C
22
Is 4-Bit Quantization More Than Noise to a Chat Model's Safety Behaviour? Attributing Behavioural Shifts with Matched Perturbations
📄 Paper arXiv cs.CR 1 ngày trước

arXiv:2611.22186v1 Announce Type: new Abstract: Chat models on phones and laptops are usually 4-bit copies of the model that was evaluated, and audits of such copies disagree about what quantization does to refusal. A 4-bit build is one of many weight perturbations of the same s…

🅲 hạng C
21
Is Abliteration Getting Harder? Not Uniformly: Refusal Removal by Direction-Based Ablation and ARA in Heretic on the Qwen3.8, Gemma 4, Olmo 3.1 and Mistral Small 3.2 LLMs Versus Their Predecessors
📄 Paper arXiv cs.AI 3 ngày trước

arXiv:2611.17846v1 Announce Type: new Abstract: Abliteration, a fine-tuning-free weight edit, removes refusal from open-weight large language models (LLMs). The developer of Arbitrary-Rank Ablation (ARA), an optimisation-based variant, reports that direction-based abliteration, …

🅲 hạng C
20
Effective Teacher Support for Reasoning Distillation
📄 Paper arXiv cs.AI 6 ngày trước

arXiv:2611.14850v1 Announce Type: new Abstract: Forward-KL language-model distillation normally computes dense teacher and student scores over the full vocabulary at every supervised position, making training and stored teacher targets expensive. We instead distill from the teac…

🅲 hạng C
20
Stress-Testing Sandbagging Detection Under Adaptive Attack
📄 Paper arXiv cs.CR 6 ngày trước

arXiv:2611.14950v1 Announce Type: new Abstract: Developers, auditors and researchers read evaluation scores as evidence of what a model can do. A model that sandbags, scoring low on purpose while keeping the capability, turns a low score into false reassurance. Sandbagging detec…

🅲 hạng C