Hard arXiv-grep of EP Patterns 3521–3544 against the full News articles blob found 24 CLEAN / 0 SKIPs. EP max observed Pattern 3544. Band arXiv 2609.28442–2609.28587.
Cumulative unique CLEAN now ≈ 2,614+ (prior ≈2,590+ through P3520 + 24).
- P3521 arXiv:2609.28442 — Order-Invariant Answers, Order-Sensitive Representations in Mathematical Reasoning
- P3522 arXiv:2609.28449 — Can LLMs Reason About Runtime Behavior? A Repository-Level Dynamic Benchmark
- P3523 arXiv:2609.28467 — Where Should I Join? Robot Group Joining via Language-Guided Goal Prediction
- P3524 arXiv:2609.28470 — StudentBench: AI and human tutoring yield equivalent GRE learning gains
- P3525 arXiv:2609.28475 — When Should Forecasting Agents Reason? Behavioral Stress Tests for Reliability Routing
- P3526 arXiv:2609.28491 — Hybrid Variational Quantum-Classical Framework with Adaptive Weighting and Efficiency Assessment
- P3527 arXiv:2609.28504 — AI in Science: Early Insights
- P3528 arXiv:2609.28506 — TW3Cast: A Frozen Router of Lightly Fine-Tuned Foundation Models for Time-Series Forecasting on GIFT-Eval
- …
- P3541 arXiv:2609.28581 — Auditability Is Not One Property: Rule Overlap, Behavioural Agreement, and Composition in Reinforcement Learning
- P3542 arXiv:2609.28582 — SGA: Uncertainty Quantification for Multi-Step Forecasting in Time Series Foundation Models
- P3543 arXiv:2609.28585 — Persistent Billable State: Denial-of-Wallet Attacks and Defenses in Tool-Calling LLM Agents
- P3544 arXiv:2609.28587 — NumericJev: Jev-like LLM Numerical Decoding with Multiway Decision Trees
Sample themes: order-invariant math answers, repo-level LLM runtime reasoning, robot group joining, StudentBench GRE tutoring parity, forecasting-agent stress tests, hybrid variational quantum, PAWS world simulation, BaseCamp DNA sequencing agents, LLM harness fingerprinting, Denial-of-Wallet billable-state attacks, NumericJev decoding.
Source EP: emerging-patterns.html · GLM advancing past 3533 · process desk.