发布日期:2026-11-15
收录条目:20
1. Cmpnd Releases dspy-wsp: Weight-Space Programming Comes to DSPy
- 来源:MarkTechPost
- 发布时间:2026-11-14 16:45 UTC
- 链接:https://www.marktechpost.com/2026/11/14/cmpnd-dspy-wsp-weight-space-programming/
摘要:Cmpnd released dspy-wsp, a prototype DSPy optimizer for weight-space programming. A frontier LLM reads a small model’s weights, edits them in a sandbox, and tests each change, with no gradients and about $5 to $10 per ru
2. Stanford Research Open-Sources Memorilla, a Latent Memory Module for Frozen LLMs
- 来源:MarkTechPost
- 发布时间:2026-11-14 10:21 UTC
- 链接:https://www.marktechpost.com/2026/11/14/stanford-memorilla-memory-system-16-memory-tokens/
摘要:Stanford researchers released Memorilla, an open-source memory system that compresses a document collection into 16 memory tokens for a frozen Qwen3-8B. It reaches 84.33% on FactKG using 108x fewer document tokens than f
3. Vercel Releases UI SDK: 50+ UI Behaviors for AI Coding Agents
- 来源:MarkTechPost
- 发布时间:2026-11-14 06:38 UTC
- 链接:https://www.marktechpost.com/2026/11/13/vercel-ui-sdk-agent-ui-behaviors/
摘要:Vercel Labs released Vercel UI SDK, an Apache-2.0 beta library of 50+ UI behaviors. Coding agents compose them into accessible components for React, Vue, Svelte, Solid and Web Components, with a bridge for direct UI cont
4. Zugwise: Domain-Aware Game Reconstruction for Digitizing Handwritten Chess Scoresheets
- 来源:arXiv cs.AI
- 发布时间:2026-11-14 05:00 UTC
- 链接:https://arxiv.org/abs/2611.14544
摘要:arXiv:2611.14544v1 Announce Type: new Abstract: Digitizing handwritten chess scoresheets has been studied as a handwriting-recognition problem: read the move in each box and check that it is legal. This framing misses th
5. Cleaning an Agent's History Hurts When the World Still Holds the Error
- 来源:arXiv cs.AI
- 发布时间:2026-11-14 05:00 UTC
- 链接:https://arxiv.org/abs/2611.14654
摘要:arXiv:2611.14654v1 Announce Type: new Abstract: Tool-using language-model agents accumulate histories of their own mistakes, and a system that compacts, repairs or hands off such a history decides what it says about thos
6. Together Compiler: Continual Compilation for AI Inference
- 来源:arXiv cs.AI
- 发布时间:2026-11-14 05:00 UTC
- 链接:https://arxiv.org/abs/2611.14681
摘要:arXiv:2611.14681v1 Announce Type: new Abstract: Today's AI workloads are rapidly changing: new models are released more often, model architectures are increasingly diverse, and traffic distributions change frequently. Th
7. When Should Semantic Priors Enter Partially-Identified Causal Discovery?
- 来源:arXiv cs.AI
- 发布时间:2026-11-14 05:00 UTC
- 链接:https://arxiv.org/abs/2611.14736
摘要:arXiv:2611.14736v1 Announce Type: new Abstract: Observational causal discovery often leaves causal di- rections unresolved or unstable. Large language mod- els (LLMs) can provide semantic information to resolve these dir
8. SynthCorp: A Platform for Creating Benchmarks for Data Agents
- 来源:arXiv cs.AI
- 发布时间:2026-11-14 05:00 UTC
- 链接:https://arxiv.org/abs/2611.14765
摘要:arXiv:2611.14765v1 Announce Type: new Abstract: Current benchmarks for LLM-driven data agents saturate quickly and cannot target specific capabilities. We present SynthCorp, a platform that creates customizable data esta
9. Rational Use of Time for Social Coordination in Humans and AI
- 来源:arXiv cs.AI
- 发布时间:2026-11-14 05:00 UTC
- 链接:https://arxiv.org/abs/2611.14766
摘要:arXiv:2611.14766v1 Announce Type: new Abstract: All social interactions are situated in time. Previous studies have shown that humans are able to use implicit temporal cues, such as the duration of a pause, to infer othe
10. Leaderboard Hacking: Preference-Based Model Evaluations are Vulnerable to Manipulation
- 来源:arXiv cs.AI
- 发布时间:2026-11-14 05:00 UTC
- 链接:https://arxiv.org/abs/2611.14769
摘要:arXiv:2611.14769v1 Announce Type: new Abstract: Arena-style rankings of language models are a widely used evaluation framework in online leaderboards, research papers, and model evaluation in production. These rankings a
11. Robust Task-Aware State Estimation from Pre-trained Perception
- 来源:arXiv cs.AI
- 发布时间:2026-11-14 05:00 UTC
- 链接:https://arxiv.org/abs/2611.14781
摘要:arXiv:2611.14781v1 Announce Type: new Abstract: Agents that perform state estimation using pretrained perception models, e.g., Vision-Language Models (VLMs), face the fundamental challenge that predictions are uncertain.
12. When Is Proof-Step Verification Worth Its Cost? Operating Requirements of Certify-or-Decline Verification for Open-Weight Mathematical Reasoning
- 来源:arXiv cs.AI
- 发布时间:2026-11-14 05:00 UTC
- 链接:https://arxiv.org/abs/2611.14794
摘要:arXiv:2611.14794v1 Announce Type: new Abstract: Proof-step verification rewrites a solver's answer into typed steps and uses LLM judges to certify or decline. We evaluate this pipeline on open-weight models across GPQA,
13. GenPFN: An In-Context Tabular Foundation Model Approach for Augmenting Small Datasets in Engineering
- 来源:arXiv cs.AI
- 发布时间:2026-11-14 05:00 UTC
- 链接:https://arxiv.org/abs/2611.14804
摘要:arXiv:2611.14804v1 Announce Type: new Abstract: Costly simulations and experiments limit the data available for engineering design. This paper proposes GenPFN, a training-free candidate generator guided by a pretrained t
14. AutoAssumption: Searching over Assumptions Discovers ML-Systems Algorithms
- 来源:arXiv cs.AI
- 发布时间:2026-11-14 05:00 UTC
- 链接:https://arxiv.org/abs/2611.14844
摘要:arXiv:2611.14844v1 Announce Type: new Abstract: Many advances in ML systems came from breaking an assumption that earlier methods took for granted: speculative decoding dropped the assumption that a model needs one forwa
15. Open Challenges and Opportunities in Human-Agent Systems
- 来源:arXiv cs.AI
- 发布时间:2026-11-14 05:00 UTC
- 链接:https://arxiv.org/abs/2611.14846
摘要:arXiv:2611.14846v1 Announce Type: new Abstract: Agentic AI systems increasingly undertake extended tasks and act on people's behalf, making human-agent interaction central to their effectiveness. Successful interaction r
16. Effective Teacher Support for Reasoning Distillation
- 来源:arXiv cs.AI
- 发布时间:2026-11-14 05:00 UTC
- 链接:https://arxiv.org/abs/2611.14850
摘要: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
17. Can an AI Agent Be Epistemically Akratic? Higher-Order Evidence & the Limits of Self-Correction
- 来源:arXiv cs.AI
- 发布时间:2026-11-14 05:00 UTC
- 链接:https://arxiv.org/abs/2611.14889
摘要:arXiv:2611.14889v1 Announce Type: new Abstract: When we judge that our reasoning was defective, it seems plausible that we should abandon the conclusion it produced. This thought has special force for AI agents built to
18. MuEdit: An efficient multi-task editing method towards inter-domain knowledge conflicts
- 来源:arXiv cs.AI
- 发布时间:2026-11-14 05:00 UTC
- 链接:https://arxiv.org/abs/2611.14910
摘要:arXiv:2611.14910v1 Announce Type: new Abstract: Large language models (LLMs) require frequent knowledge updates across heterogeneous domains, but existing editing methods fail when applied sequentially to multi-task sett
19. Edge-Native Multimodal RAG: Hardware-Aware Quantization and Architectural Optimization for Consumer GPUs
- 来源:arXiv cs.AI
- 发布时间:2026-11-14 05:00 UTC
- 链接:https://arxiv.org/abs/2611.14924
摘要:arXiv:2611.14924v1 Announce Type: new Abstract: Deployment of retrieval-augmented generation (RAG) for Edge-AI is constrained by memory and bandwidth limitations. A standard 7B parameter model consumes 14 GB Video Random
20. Provably Optimal HBT Legalization in Seconds: Certified Optima Reveal Where the Headroom Really Lies
- 来源:arXiv cs.AI
- 发布时间:2026-11-14 05:00 UTC
- 链接:https://arxiv.org/abs/2611.14928
摘要:arXiv:2611.14928v1 Announce Type: new Abstract: Hybrid bonding terminal (HBT) legalization assigns every terminal of a face-to-face 3D IC to its own pitch-grid site with minimum total Manhattan displacement. Open3DBench