- Rating
- 3
- Confidence
- 2
- Informativeness
- 2
- Presentation
- 3
- Soundness
- 3
Read review
Paper summary
The paper introduces TACIT, a guard for agent trajectories that classifies hidden representations from a frozen LLM rather than generating a safe/unsafe label. It distinguishes harmful content from unsafe tool use, showing that both are linearly decodable internally, although standard guard outputs perform near chance on controlled tool-use pairs. A regularized linear probe is trained with binary cross-entropy, while its layer and pooling method are selected through validation. TACIT outperforms released guards across six benchmarks and is competitive with full fine-tuning at substantially lower training and decoding cost.
Critical strengths
- Important distinction between message-level harmfulness and context-dependent tool-use risk.
- Controlled paired analysis and unused-tool placebo provide useful evidence.
- Simple, efficient method requiring few trainable parameters and no decoding.
- Broad evaluation, including two backbones, six benchmarks, leave-one-benchmark-out transfer, and matched-data fine-tuning.
- Intermediate-layer selection is empirically validated rather than assumed.
Critical weaknesses
- The controlled tool-use analysis covers only two schema-degradation categories and may detect schema anomalies rather than general agentic risk.
- Modified trajectories are not re-executed, so their actual harmful consequences are unclear.
- TACIT receives benchmark-specific supervision, whereas released guard baselines do not, confounding the main comparison.
- Layer 22 is dataset/model-specific, and robustness across splits, seeds, and distribution shifts is not established.
- Evaluation is mainly post-hoc; practical pre-action intervention, false-positive cost, calibration, and adversarial robustness remain untested.
Questions for authors
- How stable are the selected layer and performance across random seeds and data splits? Could the selection of layer 22 be dataset-specific?
- Can TACIT detect risk before tool execution without blocking trajectories that would end in safe refusals?
- What are its recall at low false-positive rates and robustness to adversarial schema or trajectory modifications?
3 comments
Authors Comment
Dear Reviewer aK8t,
Thank you for your review. We would appreciate hearing whether our response addresses your concerns, and are happy to clarify anything further.
Best regards,
The Authors
Thank the authors for their responses and I keep my positive score.
Authors' Rebuttal
We thank the reviewer for the careful review.
In Section 3.1, we chose the TraceSafe schema pairs because they provide matched pairs naturally, with every message and the final action byte-identical. However, they cover only how an action relates to its tool. With the same design, we built pairs for the two other relations that define unsafe tool use, from AgentDojo rollouts in which the agent carried out its actions:
Every agent action stays the same within each pair. We measure rank accuracy, averaged over the models in each row:
As controls, the same edits with content unrelated to safety, moving ordinary tool output into the request or ending the request with a sentence on reply length, give only 0.59 and 0.63, so the separation does not come from the edits as such. The internal signal is therefore not limited to schema anomalies.
The schema pairs keep TraceSafe's own labels, which mark the hazard each edit introduces. In the instruction-source pairs, the unsafe member is the exact rollout AgentDojo executed, and AgentDojo's own check confirms the agent carried out the injected task. There, the internal signal holds where the harm happened.
We fine-tuned the three guards with the same data and recipe as Table 6, and fitted TACIT on each fine-tuned guard's internal states:
With the same supervision, reading internal states beats decoding a label in all three guards. TACIT's refined readouts on the frozen backbones in Table 1, 85.1 to 86.2, also stay above every fine-tuned guard.
Figure 5 tests whether the probe depends on a single layer: layers 15 to 24 of Qwen3-4B score within about one point of each other, and layer 22 is the highest.
On the robustness concerns:
Appendix D discusses this setting. The backbone is causal, so TACIT can score every prefix of a trajectory and run while the agent acts. We trained and tested the probe on trajectories cut just before the agent's final action, where a monitor would decide whether it may run. Macro-F1 reaches 76.8 on Qwen3-4B and 78.8 on Llama-3.1-8B, and the false-positive rate rises by only about four points over complete trajectories. This points to TACIT's potential for streaming detection.
Full streaming detection also needs step-level labels, which the six benchmarks do not provide. As Appendix D notes, we leave it to future work.
We measured recall at low false-positive rates and calibration error on the pooled test splits:
At a 1% false-positive rate, TACIT recalls about ten times as many unsafe trajectories as any guard, and its calibration error is a fifth or less of every guard's. We leave adversarial robustness to future work, as Section 5.3 discusses.
We thank the reviewer again for the supportive and constructive review, which has made the paper stronger. The revised manuscript includes all of the results and changes above, and we would be glad to answer any further questions.