Probe
A lightweight classifier, typically a linear layer or small MLP, that reads intermediate-layer activation vectors as a side channel during a large model's forward pass and outputs a score for whether a target behavior, such as deceptive intent, is present. It does not modify the underlying model weights, resulting in very low deployment cost.
Timeline (last 90 days)
The probe is a lightweight classifier, usually a linear layer or a small MLP, which during training takes the activation vector of a certain layer of the model as input and outputs a score for whether the target behavior exists, without modifying the weights of the underlying large model.
The new probe architecture can aggregate information across multiple layers and multiple token positions, whereas traditional linear probes are usually trained only on a single layer or the last token.
In the SHADE-Arena evaluation, the probe reached 98.8% AUC, surpassing the Opus 5.5-based text monitoring baseline.
As the scale of the underlying model increases, the probe's detection performance instead improves, which the researchers explain as larger models having richer and more structured internal representations.
In introspection-style deception tests, the probe needs to distinguish conversation records containing the model's true hidden objective from records containing other objectives, and the probe's AUC reaches as high as 99.7%.
On multiple mainstream open-weight models, the probe can detect dishonest expression by models on politically sensitive topics as well as false statements about their own beliefs when under pressure.
The probe needs access to the model's internal activation states, so it cannot be used directly for fully black-box API services.
Current evaluations mainly focus on known types of deceptive behavior, and the probe's generalization ability to entirely new deception techniques remains to be verified.
If a model learns during training to counter probe detection by encoding deceptive intent into representation spaces that are hard for the probe to capture, detection performance may decline.
In real-world deployment, the probe is better suited as one component of a multi-layer defense system, working together with behavioral sandboxes, permission constraints, output review, and other means, rather than bearing all monitoring responsibility alone.
1 more timeline events
All Facts (11)
The probe is a lightweight classifier, usually a linear layer or a small MLP, which during training takes the activation vector of a certain layer of the model as input and outputs a score for whether the target behavior exists, without modifying the weights of the underlying large model.
50%UnverifiedThe new probe architecture can aggregate information across multiple layers and multiple token positions, whereas traditional linear probes are usually trained only on a single layer or the last token.
50%UnverifiedIn the SHADE-Arena evaluation, the probe reached 98.8% AUC, surpassing the Opus 5.5-based text monitoring baseline.
50%UnverifiedAs the scale of the underlying model increases, the probe's detection performance instead improves, which the researchers explain as larger models having richer and more structured internal representations.
50%UnverifiedIn introspection-style deception tests, the probe needs to distinguish conversation records containing the model's true hidden objective from records containing other objectives, and the probe's AUC reaches as high as 99.7%.
50%UnverifiedOn multiple mainstream open-weight models, the probe can detect dishonest expression by models on politically sensitive topics as well as false statements about their own beliefs when under pressure.
50%UnverifiedThe probe needs access to the model's internal activation states, so it cannot be used directly for fully black-box API services.
50%UnverifiedCurrent evaluations mainly focus on known types of deceptive behavior, and the probe's generalization ability to entirely new deception techniques remains to be verified.
50%UnverifiedIn real-world deployment, the probe is better suited as one component of a multi-layer defense system, working together with behavioral sandboxes, permission constraints, output review, and other means, rather than bearing all monitoring responsibility alone.
50%UnverifiedDue to their strict constraints on character count, rhyme, and tonal patterns, regulated verse poems naturally constitute a class of probe texts with complete structural annotations for use in interpretability research.
50%UnverifiedIf a model learns during training to counter probe detection by encoding deceptive intent into representation spaces that are hard for the probe to capture, detection performance may decline.
50%