Interpretability & Mechanistic Interp - Weekly Roundup

Governance-as-Code (GaC), a framework of 43 machine-checkable acceptance criteria across six compliance modules that run in a CI/CD pipeline and emit Article-indexed audit evidence, is delivered and the actual Rego policy code is shown, showing the actual Rego policy code rather than merely describing it.

These findings verify the original paper's claim that User Profile Recommendation achieves competitive performance under its test-set reranking protocol and makes recommendations more transparent.

PetriBench is introduced, a compact, fully self-contained, and scalable benchmark for evaluating LLM reasoning over dynamic state spaces using Petri nets, a mature formalism for modeling real-world concurrent and distributed systems.

Robustness controls show that the signal extends beyond English, is not explained solely by accompanying OCR, and depends on paired visual evidence, thus, routing, rather than representation alone, is a recurring bottleneck in harmful meme classification.

Governance-as-Code (GaC), a framework of 43 machine-checkable acceptance criteria across six compliance modules that run in a CI/CD pipeline and emit Article-indexed audit evidence, is delivered and the actual Rego policy code is shown, showing the actual Rego policy code rather than merely describing it.

These findings verify the original paper's claim that User Profile Recommendation achieves competitive performance under its test-set reranking protocol and makes recommendations more transparent.

PetriBench is introduced, a compact, fully self-contained, and scalable benchmark for evaluating LLM reasoning over dynamic state spaces using Petri nets, a mature formalism for modeling real-world concurrent and distributed systems.

Robustness controls show that the signal extends beyond English, is not explained solely by accompanying OCR, and depends on paired visual evidence, thus, routing, rather than representation alone, is a recurring bottleneck in harmful meme classification.

Results show that POCUS probe accuracy alone does not identify the objective that transfers best across datasets, and the ranking reverses on Mendeley-Uganda, where MoCo performs best and VideoMAE performs best.

This work introduces an in-trajectory probing framework to study intermediate answer states in agentic RAG, and forms two prediction tasks, partial answer quality prediction and partial utility prediction, and studies trajectory-derived signals from intra-iteration, inter-iteration, and query-iteration perspectives.

A layer-calibrated aggregation variant that applies per-layer adapters before pooling is evaluated, which improves the robustness of multi-layer fusion and narrows the gap to full fine-tuning while keeping the backbone frozen.