{"id":"59f44aa4-bcf9-4d90-a878-9f2f42e60f19","arxiv_id":"2611.16803","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"Under a protocol that scores alarms relative to a failure's onset, response-level AUROC is shown to hide genuine differences in early warning on repetitive degeneration, where linear probes warn before most loops on two of three models.","lead":"This paper introduces PREWARN, a protocol that evaluates token-level alarms on language model failures by when they fire relative to the failure's onset, and applies it to repetitive degeneration in three open models. It shows that response-level AUROC, the standard metric, saturates and hides large differences in whether and how early a monitor warns, and that linear probes can warn before a loop begins on two of the three models.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek/deepseek-v4-flash","headline":"The early-warning interpretation is confounded with response length because the onset usually falls after healthy responses end, and the paper acknowledges but does not control for this.","rationale":"The reader's weakest_assumption correctly identifies the length/position confound as the load-bearing concern. The paper acknowledges this limitation (b187) but does not resolve it. The headline claim that probes warn before the onset on most loops is the central empirical result, and if the confound is not controlled for, the interpretation that probes detect a signal of the coming failure is unsupported. The PREWARN protocol itself is solid, and the critique of AUROC is well-supported, but the early-warning interpretation for Apertus and Llama requires a control for length/position. The concrete test I propose would settle whether the confound actually explains the results. I agree with the reader that the verdict should remain CONDITIONAL until this confound is addressed. I do not recommend changing the verdict because the paper is honest about the limitation, and the protocol and negative results (Mistral, baselines) are still valuable. However, the headline claim should be softened or the confound controlled before the paper can be taken as establishing pre-onset signal detection.","tokens_in":37801,"tokens_out":698,"duration_ms":8115,"concrete_test":"Evaluate the frozen and adapted probes on healthy responses that are artificially truncated to the same length distribution as degenerate responses (e.g., by cropping each healthy response to the length at which the onset occurs in degenerate responses). If the probes' warning coverage and early-warning rate on these length-matched healthy responses are comparable to those on degenerate responses, then the probes are reading length/position rather than a signal of the coming loop. Alternatively, train a probe that uses only the token index (position) as input and compare its performance to the activation-based probes; if the position-only probe achieves similar warning coverage, the confound is confirmed.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The paper's central claim that linear probes warn before the onset on most degenerate responses of Apertus and Llama rests on the assumption that pre-onset flags reflect a genuine signal of the coming loop rather than response length or position. The authors explicitly state this limitation (b187): 'on APERTUSand LLAMAthe onset usually comes after a healthy response has ended, so part of what the probes flag may reflect how long the response has run (Section 4.4).' Since every degenerate response runs to the token cap (4096 tokens) while healthy responses often terminate earlier, a probe that simply flags tokens based on position or length would appear to warn before the onset on degenerate responses, without actually detecting any signal of the failure. The paper acknowledges this confound but does not control for it. Without a control—such as evaluating probes on healthy responses truncated to the same length as degenerate ones, or conditioning on response length—the headline claim that probes 'warn before the onset on most loops' does not demonstrate early knowledge of the failure. The confound is load-bearing because if the probes are largely reading length/position, then the early-warning interpretation collapses, and the paper's main empirical contribution (that probes warn before the onset) would be an artifact of the data distribution rather than evidence of pre-onset signal detection.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek/deepseek-v4-flash","summary":"The paper introduces PREWARN, a protocol for evaluating token-level monitors of language model failures that have an annotated onset. It calibrates thresholds on validation data at a fixed error budget and scores alarms relative to the onset for three tasks: response classification, failure detection, and early warning. As a case study, the authors apply PREWARN to repetitive degeneration in three open LLMs (Apertus, Llama, Mistral) using linear probes on activations and six untrained baselines. They find that response-level metrics (AUROC, accuracy) saturate and do not distinguish early-warning performance, while PREWARN reveals that probes warn before the onset on most degenerate responses of Apertus and Llama but not Mistral. Low-rank adaptation increases warning coverage but not the early-warning rate. No untrained baseline warns much more than random alarms.","tokens_in":38039,"tokens_out":2360,"duration_ms":16126,"significance":"The paper makes a clear methodological contribution by formalizing the distinction between detection and early warning for token-level monitors, and by providing a protocol (PREWARN) that is principled, practical, and applicable to any failure with an onset. The case study on repetitive degeneration is thorough, with careful human verification of labels, calibration on validation, held-out sources, and multiple seeds. The finding that response-level metrics are misleading for early-warning evaluation is important and likely to influence future work on monitoring LLMs during generation. The release of calibrated probes is a concrete asset for the community. If the results hold, the paper demonstrates that linear probes can provide early warning on some models, which has practical implications for safe deployment.","major_comments":[{"comment":"The paper acknowledges (b187) that on Apertus and Llama 'the onset usually comes after a healthy response has ended, so part of what the probes flag may reflect how long the response has run.' This is a load-bearing confound for the central claim that probes warn before the onset on most loops. Because every degenerate response runs to the token cap (4096 tokens) while healthy responses often terminate earlier, a probe that simply flags tokens based on position or length would appear to warn before the onset on degenerate responses, without actually detecting any signal of the failure. The authors acknowledge this limitation but do not control for it. Without a control—such as evaluating probes on healthy responses truncated to the same length as degenerate ones, or conditioning on response length—the headline claim that probes 'warn before the onset on most loops' does not demonstrate early knowledge of the failure. This confound is load-bearing because if the probes are largely reading length/position, then the early-warning interpretation collapses, and the paper's main empirical contribution would be an artifact of the data distribution rather than evidence of pre-onset signal detection. I consider this a major issue that needs to be addressed, either by providing such a control or by substantially tempering the claim.","section":"Section 4.4, limitation b187","authors_acknowledge":"Fourth, on APERTUSand LLAMAthe onset usually comes after a healthy response has ended, so part of what the probes flag may reflect how long the response has run (Section 4.4).","authors_acknowledge_p":0.92},{"comment":"The paper argues that response-level metrics are misleading because they rate scorers as equal that differ in early-warning performance. However, the comparison in Figure 2 and Table 4 uses the token-cap rule as a baseline, which is a straw man: it flags only the last token of a capped response, so it cannot warn. The more relevant comparison is between the frozen and adapted probes, which do differ in AUROC (0.9994 vs 0.9991 on Apertus, Table 1). While these differences are small, they are not zero, and the paper's claim that 'response-level metrics saturate' is supported by the fact that all probes have AUROC > 0.99. However, the claim that 'AUROC ranks the probes consistently, but not by how much they warn' (b111) is based on a Kendall's tau of 0.33 between validation AUROC and test warning coverage. This is a weak correlation, but the paper does not report whether the correlation is statistically significant or provide confidence intervals. Given that the number of probes is 27, a tau of 0.33 may be non-zero but still weak. The paper should report a p-value or confidence interval for this correlation to support the claim that AUROC is a poor predictor of early-warning performance.","section":"Section 4.2, Figure 2 and Table 4","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The paper reports that on Mistral, probes fire mostly after the onset and work better as detectors. This is a clear and honest result. However, the paper also claims that 'probes warn before the onset on most degenerate responses of two of the three models' (b7). On Apertus, the early-warning rate is 0.66-0.72 (Table 1), meaning about 30% of alarms come after the onset. On Llama, the early-warning rate is 0.58-0.66. So on Apertus and Llama, the probes warn before the onset on most loops, but a substantial minority (28-42%) of alarms are still after the onset. The paper should be more precise about what 'most' means and discuss the implications for deployment: a monitor that warns before the onset on only 60-70% of failures may not be reliable enough for high-stakes applications.","section":"Section 4.3, Figure 3 and Table 1","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The paper finds that low-rank adaptation more than doubles warning coverage but barely raises the early-warning rate. The interpretation is that adaptation adds flags long before the onset, which increases coverage but does not change the first alarm. However, the paper also notes that on Llama, the added flags may partly track response length (b147). This is consistent with the length confound mentioned above. The paper should explicitly test whether the added flags on Llama are correlated with response length, and if so, whether the early-warning signal is actually present or is an artifact of length.","section":"Section 4.4, Figure 4 and Table 1","authors_acknowledge":"Fourth, on APERTUSand LLAMAthe onset usually comes after a healthy response has ended, so part of what the probes flag may reflect how long the response has run (Section 4.4).","authors_acknowledge_p":0.85}],"minor_comments":[{"comment":"Equation (1) defines the threshold as the minimum value such that the false-alarm rate on validation is at most beta. However, the paper does not specify how ties are broken or what happens if no threshold achieves exactly the budget. The text says 'the lowest value at which the monitor's error rate on the validation split stays within the budget,' which is clear, but it would be helpful to note that the threshold is chosen from the set of observed scores, and that the false-alarm rate is computed as the share of healthy responses with any alarm.","section":"Section 3, Step 1","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The baselines are tuned on validation using the same selection rule as the probes (highest Warn256 at beta=1% among those with post-onset coverage >= 0.3). However, for windowed entropy and negative log-probability on Llama, no setting reaches the floor, so the one with the highest Warn256 is selected. This is a reasonable approach, but the paper should report which settings were selected for each baseline and model, as is done in Tables 6-9. The main text could briefly summarize the selected window sizes (e.g., '1,024 to 4,096 tokens' for repetition rules, '256 to 512 tokens' for confidence baselines).","section":"Section 4.1, Baselines","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"Table 1 reports that the token-cap rule has AUROC 0.9990 on Apertus, 0.9988 on Llama, and 0.9995 on Mistral. This is indeed near-perfect, but the paper should note that this is because the token-cap rule flags only the last token of a capped response, and since all degenerate responses are capped and most healthy responses are terminated, the rule perfectly separates the two classes except for the few healthy responses that are also capped. This is a trivial baseline that cannot warn, and its high AUROC is a feature of the data distribution, not of the rule's intelligence. The paper makes this point, but it could be stated more explicitly.","section":"Section 4.5, Table 1","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The conclusion states that 'probes on Apertus and Llama fire before the onset on most of the loops they flag.' This is accurate, but it would be helpful to also state the early-warning rates explicitly (66-72% on Apertus, 58-66% on Llama) to give readers a quantitative sense of 'most.'","section":"Section 5, Conclusion","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The paper lists five limitations, which is commendable. However, the fourth limitation (b187) is the most critical, as discussed in the major comments. The paper should consider adding a control experiment or at least a more detailed discussion of how the length confound could affect the results.","section":"Section 6, Limitations","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The paper reports that on Apertus, 13 of 139 onsets are off by 256 tokens or more, and on 10 of these the judge places the onset 380-1250 tokens after the reviewer does. This is a significant discrepancy that could affect the early-warning metrics. The paper should discuss whether these cases are concentrated in particular sources or response types, and whether the human-verified labels are considered the ground truth for all analyses.","section":"Appendix A.2, Human verification","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"Table 4 shows that shuffling scores within a response produces a scorer with the same AUROC but very different early-warning performance. This is a useful sanity check, but the paper should note that shuffling is not a realistic monitor because it uses future information. The point is to illustrate that AUROC cannot distinguish between a monitor that fires early and one that fires late, which is well-taken.","section":"Appendix A.5, Table 4","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The paper is well-written and the figures are clear. The use of multiple seeds, human verification, and held-out sources strengthens the empirical contribution. The code and model releases are valuable.","section":"General","authors_acknowledge":null,"authors_acknowledge_p":null}],"recommendation":"major_revision"},"author_rebuttal":null,"desk_editor":null,"pith_extraction":null,"created_at":"2026-11-13T12:34:47.303974+00:00","model_set":{"reader":"deepseek/deepseek-v4-flash"},"pipeline_version":"pith-pipeline@v1.4.1-daily-deepseek-or-medium","input_kind":null,"input_chars":null,"input_total_chars":null,"passes_missing":null,"published_standing":null,"published_confidence":null,"objection_audit":null,"run_record":null,"basis":{"kind":"not_recorded","source":null,"chars_given":null,"chars_total":null},"receipt":null,"falsifier":null,"supporting_citations":[],"lane_accuracy":{"status":"not yet measured","lane":{"model":"deepseek/deepseek-v4-flash","input_kind":null},"url":"https://pith.science/accuracy"},"review_version":1}