- Rating
- 2
- Confidence
- 3
- Informativeness
- 1
- Presentation
- 3
- Soundness
- 2
Read review
Paper summary
This paper studies trajectory-level safety auditing for long-horizon LLM agents, where unsafe behavior may depend on sparse, delayed, or compositional evidence distributed across multiple interactions. It proposes Global Auditing via Evidence Latents (GAVEL), a two-pass framework comprising an Evidence Encoder and an Audit Decoder. The encoder processes the complete observed trajectory and produces compact evidence representations using 16 learnable queries, while the decoder assesses the original trajectory conditioned on these representations. Training uses trajectory-level safety labels and updates the queries, encoder LoRA adapters, and classification head while keeping the pretrained backbone weights frozen. Across three safety benchmarks and four backbones, GAVEL achieves the highest mean accuracy in all 12 main evaluation settings, with improvements of up to 12.6 percentage points over the strongest competing baseline. Additional LongSafety diagnostics examine performance under dispersed evidence and increasing context lengths, while ablations and latent interventions investigate the contribution of the learned evidence context.
Critical strengths
The method is well suited to long-horizon safety auditing. GAVEL first uses an Evidence Encoder to integrate scattered evidence across the full trajectory. The Audit Decoder then makes predictions based on both the evidence representation and the original trajectory, capturing global information dependencies while preserving fine-grained behavioral details. This design is well suited to risks that emerge only after a delay or arise from combinations of multiple steps (Secs. 3.2–3.3; Fig. 2).
The training scheme is parameter-efficient and requires no fine-grained evidence annotations. The method is trained end-to-end using only trajectory-level safety labels. It updates the learnable queries, the Encoder’s LoRA parameters, and the classification head while keeping the pretrained backbone frozen. No additional annotations of evidence spans or explanatory text are required, reducing both the number of trainable parameters and annotation requirements (Sec. 3.4; Appendix B.2).
The main experiments and targeted diagnostics provide substantial evidence of effectiveness. GAVEL achieves the highest average accuracy across all 12 main experimental settings, spanning three safety benchmarks and four backbone models. In the LongSafety diagnostics, it outperforms the strongest baseline by 17.1 and 18.0 percentage points on the delayed-evidence and compositional-evidence categories, respectively, supporting its ability to handle safety evidence distributed across multiple steps (Table 1; Appendix G, Table 9).
Critical weaknesses
Although the main experiments use four different backbone models, the paper does not clearly specify which models serve as the trained Encoder and Decoder in each experimental setting.
One important ablation is still missing: keeping the rest of the framework unchanged, removing only the 16 queries appended to the Encoder input, and retraining the model.
Figure 7 uses an exponentially spaced sweep to reveal the broad trend, but this is insufficient to establish whether 16 is locally optimal.
Questions for authors
Why is LoRA used for training? Would full-parameter fine-tuning be feasible, and how would it compare?
How exactly are the evaluation metrics computed for each of the three benchmarks? The paper does not explain this in sufficient detail.
Could the authors add a baseline that directly classifies the Evidence Encoder’s output under the same training setup, to verify the necessity of the Audit Decoder?
3 comments
W3. Fine-Grained Latent Budget Sensitivity
We thank the reviewer for this constructive suggestion. While the exponential sweep () demonstrated broad stability across a 32-fold capacity range, we agree that it did not adequately characterize the local behavior around .
To investigate local sensitivity, we conducted a fine-grained sweep evaluating on Qwen3-8B across all three benchmarks under identical protocols. The table also includes the existing results from the original exponential sweep.
Table R3-3. Fine-grained latent budget sensitivity on Qwen3-8B (Accuracy in %).
| Latent Budget K | ASSEBench Acc (%) | Pre-Ex-Bench Acc (%) | R-Judge Acc (%) |
|---|---|---|---|
| 90.70 | 87.90 | 90.20 | |
| 91.08 | 91.31 | 92.76 | |
| 91.66 | 91.82 | 93.18 | |
| (Default) | 91.57 | 92.06 | 93.40 |
| 91.49 | 91.87 | 93.46 | |
| 91.21 | 91.55 | 93.12 | |
| 90.60 | 87.60 | 92.80 | |
| 90.80 | 85.40 | 93.50 | |
| 90.90 | 92.20 | 91.40 |
The fine-grained results indicate that performance remains stable around , forming a consistent high-performance plateau across , with all accuracy results within 0.24 pp of the result. Under , the model achieves 91.57% on ASSEBench, 92.06% on Pre-Ex-Bench, and 93.40% on R-Judge, sitting squarely within this optimal region. Smaller budgets () exhibit noticeable drops due to evidence bottlenecking, whereas larger budgets () show non-monotonic behavior on denser benchmarks like Pre-Ex-Bench, consistent with the Information Bottleneck perspective discussed in Appendix A.6.
We emphasize that is not claimed to be a strict global mathematical optimum across every individual dataset—for instance, Pre-Ex-Bench exhibits a secondary peak at . Rather, serves as a robust, universally effective operating point that balances compression quality and computation across distinct benchmarks.
We have incorporated Table R3-3 and this discussion into Appendix D.1.
Q1. LoRA versus Full-Parameter Fine-Tuning
We adopted LoRA primarily to achieve parameter efficiency and prevent the degradation of foundational linguistic reasoning under trajectory-level binary supervision.
To empirically compare LoRA against Full Fine-Tuning (Full FT), we trained a variant that unfreezes the full Evidence Encoder backbone (~8.0B parameters for Qwen3-8B), while keeping all other components (Audit Decoder frozen, data, learning rate schedule, epochs, batch size) strictly identical.
Table R3-4. LoRA versus full fine-tuning of the Evidence Encoder on Qwen3-8B Pre-Ex-Bench.
| Training Strategy | Trainable Params | Accuracy (%) | F1 (%) | Unsafe Recall (%) | Relative GPU Hours |
|---|---|---|---|---|---|
| LoRA (GAVEL) | 8.46M | 92.06 ± 2.26 | 89.69 ± 1.63 | 89.27 ± 1.92 | 1.00× |
| Full FT (Encoder) | ~8.00B | 92.28 ± 1.87 | 90.08 ± 1.54 | 89.91 ± 1.73 | 2.47× |
Full fine-tuning yields only marginal improvements (+0.22 pp accuracy and +0.39 pp F1) at the expense of updating ~1000× more parameters and incurring 2.47× higher training compute. This demonstrates that full parameter deformation is unnecessary; adapting attention projections via LoRA is sufficient to steer the queries toward trajectory-level risk aggregation.
We have incorporated this comparative study into Appendix B.2.
Q2. Evaluation Metrics Across the Three Benchmarks
We appreciate the reviewer noting the need for a more detailed protocol. In the revised paper, we have expanded Section 4.1 and Appendix B.1 to clarify the prediction setups and metric computations:
-
Benchmark Task Formats:
- ASSEBench: Evaluates complete multi-turn interaction traces against official safety labels under the lenient standard.
- Pre-Ex-Bench: Judges whether a proposed action in context poses potential hazards before execution.
- R-Judge: Assesses complete agent-environment records for cumulative safety violations.
-
Metric Definitions: In all benchmarks, Unsafe is designated as the positive class (), and Safe as the negative class (). Let denote true positives, false positives, true negatives, and false negatives:
-
Multi-Seed Aggregation: For each random seed, predictions are thresholded at on the test set to compute confusion matrices, reporting the empirical mean and standard deviation across seeds ().
Q3. Direct Classification from Evidence Encoder Outputs
We agree that directly evaluating whether the Evidence Encoder alone can perform safety classification helps isolate the architectural benefit of the Audit Decoder.
To verify this, we evaluated two architectural baselines under identical setups on Qwen3-8B (Pre-Ex-Bench):
- Encoder-only Classifier: Removes the Audit Decoder entirely. We apply mean-pooling over the 16 latent representations from the Evidence Encoder and feed them into a linear classification head.
- Single-pass LoRA Classifier: Employs a single backbone with LoRA to directly predict trajectory safety from the raw input in a single pass.
Table R3-5. Architectural comparisons on Qwen3-8B Pre-Ex-Bench.
| Method | Passes | Evidence Encoder | Audit Decoder | Accuracy (%) | F1 (%) | Unsafe Recall (%) |
|---|---|---|---|---|---|---|
| Encoder-only Classifier | 1 | Yes | No | 87.91 ± 1.46 | 84.03 ± 1.74 | 82.37 ± 2.02 |
| Single-pass LoRA | 1 | No | Yes | 89.37 ± 1.62 | 86.29 ± 1.85 | 84.81 ± 2.14 |
| Full GAVEL | 2 | Yes | Yes | 92.06 ± 2.26 | 89.69 ± 1.63 | 89.27 ± 1.92 |
Directly classifying from compressed evidence latents (Encoder-only) leads to a 4.15 pp drop in accuracy and a 6.90 pp drop in unsafe recall. While the latents capture global dependency cues, severe compression into 16 slots inevitably discards fine-grained execution details (e.g., specific parameters or code snippets). Conversely, a single-pass model struggles to integrate distant evidence across long contexts. GAVEL succeeds because the Audit Decoder combines global latent guidance with unconstrained causal attention back into the exact original record.
We have added these comparisons and the corresponding analysis to Section 4.4.
We sincerely thank the reviewer for the thorough and constructive evaluation, as well as the positive assessment of our two-pass framework, parameter-efficient training strategy, and empirical gains across both benchmark and diagnostic evaluations.
Below, we address the weaknesses and questions point-by-point, with corresponding new ablation experiments and clarifications incorporated into the revised manuscript.
W1. Clarification of Encoder–Decoder Configurations
We appreciate the reviewer pointing this out. In the original submission, while we stated the backbone families used, the exact instantiation and parameter relationship between the Evidence Encoder and Audit Decoder were not explicitly tabulated.
To resolve this ambiguity, we have updated Appendix B.2 with an explicit configuration table (Table R3-1). In each experimental setting:
Qwen3-8B).Table R3-1. Backbone configurations of the Evidence Encoder and Audit Decoder.
This structural decoupling ensures that global evidence organization and fine-grained verification are handled by specialized pathways while preserving the foundational knowledge of the underlying LLM backbones.
W2. Retraining Ablation for Learnable Queries
We fully agree with the reviewer. The original intervention experiments in Section 4.4 and Figure 4 primarily evaluated inference-time substitutions, which do not fully isolate the learning contribution of the queries during training.
To address this, we conducted retraining experiments from scratch on Qwen3-8B (Pre-Ex-Bench) under identical optimization setups:
Table R3-2. Retraining ablations of learnable queries on Qwen3-8B Pre-Ex-Bench.
The results show that removing learnable queries degrades accuracy by 3.64 pp and unsafe recall by 6.36 pp, demonstrating that simply reading the raw sequence hidden states is insufficient for trajectory-wide risk integration. While fixed random queries (Random-Q) partially mitigate this drop by providing decoupled readout slots, they still fall 2.89 pp short of Full GAVEL. This confirms that the observed improvements do not merely stem from extra token positions, but rely on task-driven supervision adapting the queries to bind dispersed risk evidence.
We have added this retraining ablation to Section 4.4 in the revised paper.