arXiv AI August 13, 2026 By Parvel Gu

Detecting a Route Flip Is Easier Than Knowing Whether to Fix It: Causal Route-Mediated Damage in Quantized Mixture-of-Experts

arXiv:2608. 11212v1 Announce Type: new Abstract: Top-k Mixture-of-Experts (MoE) routing is discontinuous, so a deployment-motivated numerical disturbance -- simulated 4-bit KV-cache quantization read by a protected BF16 gate -- pushes tokens across decision boundaries and flips which experts fire.

Summary generated by The Flow from the publisher's feed. The full article lives at arXiv AI.

efficiency

arXiv Machine Learning

2d ago

The Integer Alibi: Localizing Cross-Kernel Divergence in INT8-Quantized LLM Inference

arXiv:2608. 13756v1 Announce Type: new Abstract: Two GPU kernels implementing the same scaled INT8 GEMM interface are usually treated as interchangeable.

By Teng-Ruei Chen

llms efficiency benchmarks

arXiv Machine Learning

6d ago

Located but Not Releasable: Silent Gate Inversion and Bounded Linear Release

arXiv:2608. 11822v1 Announce Type: cross Abstract: A growing body of work reports that language models represent task-relevant latent structure that they fail to use.

By Xining Xun

llms

arXiv AI

1d ago

Routing Divergence Is Not Evidence of Behavioral Influence in Same-Weight MoE Self-Distillation

arXiv:2608. 15787v1 Announce Type: cross Abstract: Two Mixture-of-Experts (MoE) forward passes can share every weight yet route the same token through different experts.

By Cedric Caruzzo, Donggeun Yoo, Tae Soo Kim

efficiency

arXiv Machine Learning

Aug 5

Sensitivity, Causality, and Repair Dissociate: A Layer-Wise Analysis of Perturbation Robustness and Its Scaling

arXiv:2608. 03842v1 Announce Type: cross Abstract: When a language model fails on surface-perturbed input (typos, OCR noise, homophones), "which layer is responsible" has three natural operationalizations: where representations diverge most (sensitivity), where restoring clean activations recovers the prediction (causality), and where a small adapter can repair the damage (compensatory capacity) - and we show these three layer maps dissociate.

By Nathan Labiosa, David Buff, Ena Nayak, Erica Donno

llms computer-vision

arXiv Machine Learning

Jun 10

Quality Is Not a Safety Proxy Under Quantization

arXiv:2606. 10154v1 Announce Type: new Abstract: Quantized checkpoints are often screened first with quality metrics and only later, if at all, with direct safety tests.

By Sahil Kadadekar

llms efficiency safety

arXiv Machine Learning

Jun 10

From Observation to Intervention: A Causal Audit of Expert Importance in Mixture-of-Experts Models

arXiv:2606. 10703v1 Announce Type: new Abstract: Interpretability methods routinely use population-level summary statistics over observed model behaviour to license claims about the effects of targeted interventions on specific computations; in Pearl's terms, they treat rung-1 associational evidence as if it supported rung-2 interventional conclusions, a move whose validity is rarely tested.

By Leonard Engmann, Christian Medeiros Adriano, Holger Giese

efficiency safety

Related stories

The Integer Alibi: Localizing Cross-Kernel Divergence in INT8-Quantized LLM Inference

Located but Not Releasable: Silent Gate Inversion and Bounded Linear Release

Routing Divergence Is Not Evidence of Behavioral Influence in Same-Weight MoE Self-Distillation

Sensitivity, Causality, and Repair Dissociate: A Layer-Wise Analysis of Perturbation Robustness and Its Scaling

Quality Is Not a Safety Proxy Under Quantization

From Observation to Intervention: A Causal Audit of Expert Importance in Mixture-of-Experts Models