PAC-CF introduces a method called Probably Approximately Correct Conformal Filtering to calibrate irreversible frontier pruning in large language model (LLM)-guided search. The approach treats tree pruning as a PAC‑guaranteed decision problem, deriving a conformal margin from score deficits on held‑out Native traces and applying a direct score‑gap filtering rule during deployment. Experiments across multiple domains and controllers show that PAC‑CF improves utility at various budgets and reduces workload metrics, notably boosting equal‑domain utility by 4.38 points and cutting physical requests by 18.95% with a 23.76% token reduction on a pruning‑aware ToolTree under a 100‑request budget.
arXiv:2608.29765v1 Announce Type: new Abstract: Structured pruning uses surrogate objectives because direct task evaluation over every feasible mask is too expensive. Most evaluations report average...
arXiv:2607. 27083v1 Announce Type: new Abstract: As LLM agents increasingly depend on diverse external services such as search engines, databases, and connectors, agent harnesses face a fundamental tool-selection challenge: acquiring too few tools leaves the task under-informed, while too many adds cost, context load, and privacy exposure.
The paper introduces HOPE, a second‑order pruning method for Mixture‑of‑Experts language models that accounts for cooperative interactions between experts. Unlike first‑order methods such as REAP, HOPE derives an objective that provably bounds pruning error and is shown to outperform baselines across three large MoE models, multiple calibration sets, and diverse benchmarks, especially at high pruning rates and on agentic tasks. The results demonstrate that preserving expert interactions allows aggressive compression with minimal performance loss on complex workloads.
CANOPY is a multi‑fidelity tree bandit algorithm that learns where a piecewise‑smooth prior holds instead of assuming global smoothness. It uses cheap random‑path probes to certify local aggregation bias and then focuses expensive leaf evaluations on cells where smoothness is violated. The method achieves provable fixed‑budget and regret guarantees that scale with the number of discontinuities, matching smooth‑tree rates when no violations exist and approaching structure‑blind search when violations are dense.
Mixture-of-Experts (MoE) language models suffer from large parameter counts, which create a significant memory bottleneck. Expert pruning is the most direct approach for reducing this parameter count,...
arXiv:2603. 18492v3 Announce Type: replace Abstract: Mixture-of-Experts (MoE) language models increase parameter capacity without proportional per-token computation, yet deployment still requires storing the full expert pool, making expert pruning important for reducing memory and serving overhead.
arXiv:2608. 12953v1 Announce Type: cross Abstract: Structured pruning is a promising approach for compressing large language models (LLMs), yet existing methods rely heavily on greedy heuristics that produce myopic decisions, and often fail to precisely meet target compression budgets.
arXiv:2609.30465v1 Announce Type: cross Abstract: Mixture-of-experts (MoE) models activate few experts per token but store the full expert pool. Expert pruning reduces this storage burden; at a fixed...
arXiv:2606. 09886v1 Announce Type: cross Abstract: Sparse Mixture-of-Experts (MoE) large language models achieve strong quality with low per-token compute, yet their deployment is often limited by the memory wall: the full expert pool must remain resident to support token-dependent routing.
arXiv:2608. 01023v1 Announce Type: new Abstract: We present Caliber, an output-perturbation defense against model extraction that formulates noise selection as a calibration problem: how much the defense degrades the supervision signal used to train a surrogate, and the provable per-input query cost of recovering the clean logits.
arXiv:2606. 04661v1 Announce Type: cross Abstract: Prompts tuned for accuracy often grow long, raising inference cost on every model call.
arXiv:2607. 24647v1 Announce Type: new Abstract: AI-driven autonomous research (AR) systems are becoming increasingly effective across a broad range of tasks.