← Back to all news
arXiv Machine Learning

Calibrate-Then-Delegate: Safety Monitoring with Risk and Budget Guarantees via Model Cascades

Related stories

Hugging Face Trending Papers
Sep 2

ObserverBench: Testing Mechanistic Estimates for Intervention and Control

ObserverBench is a benchmark framework that evaluates whether internal mechanistic estimators—called observers—are suitable for guiding interventions, control, or safety actions in language models. It separates estimation accuracy from the loss incurred by the chosen action, demonstrating that accurate average estimates can still lead to poor decisions. Experiments on GPT‑2‑small, Qwen2.5‑7B, Gemma‑2‑9B‑it, and Qwen3.5‑9B show that observers trained on action loss tend to select lower‑loss actions, while traditional metrics like AUROC can rank monitors differently from deployment loss, highlighting the need for task‑specific evaluation.

llmsbenchmarkssafety
More like this →
arXiv AI
Sep 4

ObserverBench: Testing Mechanistic Estimates for Intervention and Control

ObserverBench is a benchmark framework that evaluates whether internal mechanistic estimators—called observers—are suitable for guiding interventions, control, or safety actions in language models. It separates estimation accuracy from the loss incurred by the chosen action, showing that accurate predictions do not always lead to better decisions. Experiments on GPT‑2‑small, Qwen2.5‑7B, Gemma‑2‑9B‑it, and Qwen3.5‑9B demonstrate that observers trained on action loss can reduce deployment loss, while traditional metrics like AUROC may rank monitors differently from actual performance.

By Vijay Erramilli
llmsbenchmarkssafety
More like this →
arXiv Machine Learning
Oct 23

Scaling Laws for Safety Probes

The paper introduces two scaling laws that predict the performance of safety probes—tools used to monitor language models for harmful behavior—based on model size and the number of labeled examples. By creating six large safety datasets and training probes across multiple model families, the authors show that probe error decreases as a power law with model size and further improves with more examples, with the improvement rate increasing for larger models. These laws enable researchers to forecast probe accuracy and decide whether to invest in more data or a larger model to achieve a desired error threshold.

By Yuchen Wu (University of Washington), Jerry Wei (Anthropic)
llmssafety
More like this →
Hugging Face Trending Papers
Aug 18

Judge, Retrieve, or Abstain: Uncertainty-Guarded LLM Judging with Provable Risk Guarantees

The paper introduces a risk‑controlled framework for using large language models (LLMs) as judges in tasks without reference answers. By calibrating uncertainty thresholds on a held‑out set, the method ensures that the false discovery rate of accepted verdicts stays below a user‑specified level α with high probability, using finite‑sample Clopper–Pearson intervals. When the parametric judge lacks confidence, the instance is routed to a retrieval‑augmented mode with a second calibrated threshold, preserving the error guarantee while achieving higher coverage than single‑mode baselines.

llmsragbenchmarkssafety
More like this →
arXiv Machine Learning
Sep 30

RAISE: Diagnosing Acquisition Collapse in Costly LLM Signals

The paper introduces RAISE, a diagnostic framework that tests whether a costly large language model (LLM) signal provides enough pre-call information to justify selective use. It identifies the failure mode of acquisition collapse, where an LLM appears useful overall but lacks actionable evidence for individual decisions. The authors demonstrate RAISE with Structured Hypothesis Embeddings (SHE) and evaluate it across multiple study designs, showing that predictable incremental benefit, rather than average lift, indicates recoverable selective value.

By Ying Yuan, Yu Wang, Yize Cheng, Xuyang Wu
llmsragagentsrobotics
More like this →