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

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