Capability Watch
Papers
Every frontier-AI research paper we’ve surfaced from arXiv, newest first — the raw capability signals Frontier Watch tracks. Scroll to load older ones.
1,802 papers tracked · ← dashboard
- 1Learning structural balance of graphs from quantum spectral featuresarXiv cs.AI / cs.LG / cs.CL (recent) · 2026-09-10
- 2ORCH: Organizational Principles Enable Collective Intelligence in Embodied AIarXiv cs.AI / cs.LG / cs.CL (recent) · 2026-09-10
- 3LOCUS: Task-Aware Low-Rank Post-Training for Token-Efficient Language GenerationarXiv cs.AI / cs.LG / cs.CL (recent) · 2026-09-10
- 4Building py-kvcache: A Performance Characterization of External KV Caching for vLLM with NVMe SSDsarXiv cs.AI / cs.LG / cs.CL (recent) · 2026-09-10
- 5SIRF: A Spec-Internalized Risk Foundation Model for Industrial Content Risk ControlarXiv cs.AI / cs.LG / cs.CL (recent) · 2026-09-10
- 6RAG-Safety-Bench: Reliable Evaluation of Retrieval-Augmented LLM SafetyarXiv cs.AI / cs.LG / cs.CL (recent) · 2026-09-10
- 7Component-Aware Differential Privacy for Federated Multilingual Speech-LLMsarXiv cs.AI / cs.LG / cs.CL (recent) · 2026-09-10
- 8A Unified Per-Token Gating Family for On-Policy Distillation: FKL/RKL Mixing with Multi-Channel and Bias CoefficientsarXiv cs.AI / cs.LG / cs.CL (recent) · 2026-09-10
- 9Recognizing Is Not Reversing: A Controlled Inversion Test of Fact-Preserving News FramingarXiv cs.AI / cs.LG / cs.CL (recent) · 2026-09-10
- 10Predicting Privacy Leakage from Weight Spectral DensityarXiv cs.AI / cs.LG / cs.CL (recent) · 2026-09-10
- 11SpecGuard: Inference-Time Backdoor Detection For FreearXiv cs.AI / cs.LG / cs.CL (recent) · 2026-09-10
- 12Thinking with Looped FlowsarXiv cs.AI / cs.LG / cs.CL (recent) · 2026-09-10
- 13Logit Refiner: Improving Visual Autoregressive Models via Intra-Scale Dependency ModelingarXiv cs.AI / cs.LG / cs.CL (recent) · 2026-09-10
- 14Near-Optimal Reinforcement Learning with Multi-Step Transition LookaheadarXiv cs.AI / cs.LG / cs.CL (recent) · 2026-09-10
- 15Model-Aware Schedules Improve Generation via Fiberwise Optimal TransportarXiv cs.AI / cs.LG / cs.CL (recent) · 2026-09-10
- 16From Parameters to Answers: How LLMs Retrieve and Use Their Internal KnowledgearXiv cs.AI / cs.LG / cs.CL (recent) · 2026-09-10
- 17Explainability Assistant: A Conversational XAI Interface for Interpreting Energy Consumption ModelsarXiv cs.AI / cs.LG / cs.CL (recent) · 2026-09-10
- 18RetroThinker: Enabling Retrospective Thinking in Speech LLMsarXiv cs.AI / cs.LG / cs.CL (recent) · 2026-09-10
- 19AdamX: Cosine similarity meets gradient descentarXiv cs.AI / cs.LG / cs.CL (recent) · 2026-09-10
- 20Augustinian BabyLM: What Ostensive Definition Can and Cannot Teach a Small Language ModelarXiv cs.AI / cs.LG / cs.CL (recent) · 2026-09-10
- 21The Last AI Built by Humans: Toward Genuine Recursive Self-ImprovementarXiv cs.AI / cs.LG / cs.CL (recent) · 2026-09-10
- 22On the Regularization Landscape for the Linear Recommendation ModelsarXiv cs.AI / cs.LG / cs.CL (recent) · 2026-09-10
- 23Domain-Specific Hallucination Detection in Large Language ModelsarXiv cs.AI / cs.LG / cs.CL (recent) · 2026-09-10
- 24CoRA-NAS: Coarse Ranking and Anchor-Residual Refinement for Neural Architecture SearcharXiv cs.AI / cs.LG / cs.CL (recent) · 2026-09-10
- 25Nuha-Speech: Building General-Purpose Arabic Speech-LLMsarXiv cs.AI / cs.LG / cs.CL (recent) · 2026-09-10
- 26CausalArena: Benchmarking Causal Discovery in the Foundation Model EraarXiv cs.AI / cs.LG / cs.CL (recent) · 2026-09-10
- 27MindTopo: Can Foundation Models Reason in Topological Space?arXiv cs.AI / cs.LG / cs.CL (recent) · 2026-09-10
- 28Artificial Id: Drive and Persistent Alignment in Agentic AIarXiv cs.AI / cs.LG / cs.CL (recent) · 2026-09-10
- 29Distance generalization in transformers: why bother with positional encoding?arXiv cs.AI / cs.LG / cs.CL (recent) · 2026-09-10
- 30Data Scarcity and Model Sparsity: Mixtures-of-Experts Overfit More to Repeated DataarXiv cs.AI / cs.LG / cs.CL (recent) · 2026-09-10
- 31General Quantification of Covariate and Concept ShiftsarXiv cs.AI / cs.LG / cs.CL (recent) · 2026-09-10
- 32GPU-CFR: 80x Faster Counterfactual Regret Minimization by Compiling the Game to Static Dataflow and CUDA Graph ReplayarXiv cs.AI / cs.LG / cs.CL (recent) · 2026-09-10
- 33What Should an Agent Forget? Separating What Is Stored from What Is UsedarXiv cs.AI / cs.LG / cs.CL (recent) · 2026-09-09
- 34Maverick: Private and Verifiable LLM Inference Made Practical via Matrix-Vector Multiplication DelegationarXiv cs.AI / cs.LG / cs.CL (recent) · 2026-09-09
- 35KVShareArena: KV-Cache Reuse Across Contexts and Model CheckpointsarXiv cs.AI / cs.LG / cs.CL (recent) · 2026-09-09
- 36Training Trajectories Determine Circuit Removability in Annealable Soft-Prior TransformersarXiv cs.AI / cs.LG / cs.CL (recent) · 2026-09-09
- 37GANDR: Claim Auditing for Verifiable Legal Answer GenerationarXiv cs.AI / cs.LG / cs.CL (recent) · 2026-09-09
- 38A Dominant Diffuse Phase in the Sparse Autoencoder Phase DiagramarXiv cs.AI / cs.LG / cs.CL (recent) · 2026-09-09
- 39RiLM: Parameter-Efficient Language Modeling via Geodesic DecodingarXiv cs.AI / cs.LG / cs.CL (recent) · 2026-09-09
- 40One Loop, Two Gains: Can Active Learning win the Lottery for Free?arXiv cs.AI / cs.LG / cs.CL (recent) · 2026-09-09