- Rating
- 2
- Confidence
- 2
- Informativeness
- 2
- Presentation
- 3
- Soundness
- 2
Read review
Paper summary
This paper identifies and analyses a non-termination problem in training-free soft reasoning for large language models. The authors identify that after a reasoning impasse, worsening persistent semantic recurrence leads to deep lock-in. Based on this observation, they propose E-TRACE, a training-free event-level framework that detects persistent non-local recurrence. Once a loop is detected, the method rolls back to a checkpoint outside the recurrent region and switches to hard-token decoding. Experiments on mathematical reasoning, code generation, and knowledge-intensive question answering show that the method generally improves reasoning accuracy while shortening the reasoning length.
Critical strengths
This paper addresses a practically relevant problem in soft reasoning: the tendency of models to enter infinite loops. Similar phenomena are often observed in practice with activation-based methods, potentially due to a mismatch between the reasoning process at inference time and the model's training mechanism. The paper analyses this issue at the event level and proposes a practical strategy to detect and mitigate such recurrent patterns.
Critical weaknesses
-
The analysis could be more thorough. For example, Figure 1 presents a toy example, but the paper does not adequately clarify how frequently this issue occurs, does it affect nearly all reasoning tasks, or only a subset? Similarly, it remains unclear whether the S1-S3 looping patterns are task-specific or reflect more general behaviors. Addressing these questions would provide a more comprehensive understanding of the problem and strengthen the paper.
-
When a looping pattern is detected, the method falls back to hard decoding. While this is a reasonable design choice, it is not particularly surprising. I would have preferred an approach that attempts to correct the looping behavior within the soft reasoning process itself, rather than switching back to hard decoding. That said, this could be an interesting direction for future work.
-
The main experiments do not provide strong evidence of substantial accuracy improvements. The largest gains are reported on AIME24 and AIME25, but each benchmark contains only 30 problems. This also relates to the first concern regarding how prevalent the looping problem actually is. The more significant advantage of the proposed method may instead lie in reducing reasoning length. However, unless I have overlooked something, the paper does not clearly define how reasoning length is measured. Since the method switches to hard decoding after detecting a loop, it is unclear whether the tokens generated during the looping phase are included in the reported reasoning length. These tokens still consume computational resources and time, so excluding them could raise concerns about the fairness of the comparison.
Questions for authors
-
Could the authors report how frequently this issue occurs, either across all evaluated tasks or on representative benchmarks?
-
Could the main experiments quantify how many problems are correctly solved specifically because the proposed method successfully escapes a loop?
-
How exactly is the reasoning length (#L) calculated? Does it include tokens generated during the looping phase, or only those generated in the final reasoning process?
-
Line 352: Manually inserting the </think> token may result in an incomplete thinking process. Could the authors clarify whether and how this affects subsequent hard decoding?
-
Minor issue: Lines 488-489 appear to contain an unintended line break.
12 comments
Response to Reviewer rgHN (2)
Q2 & W3 Details of the problems being solved
Thank you for this suggestion. To demonstrate E-TRACE’s ability to handle loop conditions, we report paired answer-correctness outcomes on Qwen3-8B across all eight benchmarks. The Soft Thinking baseline includes Cold Stop, which also stops early. The table distinguishes recovered cases (E-TRACE correct, Soft Thinking wrong) from reverse changes (Soft Thinking correct, E-TRACE wrong).
| Task | data size | Both Correct | Only E-Trace | Only Soft Thinking | Both wrong | Acc of E-Trace(%) | Acc of soft thinking(%) |
|---|---|---|---|---|---|---|---|
| aime2024 | 30 | 17 | 6 | 1 | 6 | 70.67 | 60.00 |
| aime2025 | 30 | 18 | 2 | 2 | 8 | 66.67 | 66.67 |
| gsm8k | 1319 | 1242 | 12 | 2 | 63 | 95.07 | 94.31 |
| math500 | 500 | 480 | 5 | 5 | 10 | 97.00 | 97.00 |
| gpqa | 198 | 91 | 30 | 14 | 63 | 61.11 | 53.03 |
| medqa | 300 | 222 | 22 | 13 | 43 | 81.33 | 78.33 |
| humanevalplus | 164 | 140 | 7 | 5 | 12 | 89.63 | 88.41 |
| mbppplus | 378 | 294 | 13 | 9 | 62 | 81.22 | 80.16 |
E-TRACE's advantage is particularly evident on tasks that remain challenging for Qwen3-8B. On GPQA, it recovers 30 problems that Soft Thinking answers incorrectly, raising accuracy from 53.03% to 61.11%. Accuracy also improves from 78.33% to 81.33% on MedQA and from 60.00% to 76.67% on AIME2024. These results demonstrate that loop-aware intervention can improve answer quality on challenging tasks.
We additionally evaluated Phi-4-Reasoning, and this model exhibits more severe looping phenomena under soft token inference. To illustrate this situation, we report the generation length. As you can see, the growth length value for Soft Thinking has become very high, which means Soft Thinking can no longer fully find a place to exit thinking merely by relying on continuous low entropy. Furthermore, E-TRACE achieves higher accuracy and shorter generation length than Soft Thinking on all eight benchmarks:
| Benchmark | Soft Thinking Acc. (%) | E-TRACE Acc. (%) | Soft Thinking #L | E-TRACE #L | Length reduction (%) |
|---|---|---|---|---|---|
| AIME2024 | 53.33 | 76.67 | 11,451.27 | 9,914.37 | 13.42 |
| AIME2025 | 46.67 | 56.67 | 11,221.53 | 8,595.40 | 23.40 |
| GSM8K | 91.96 | 94.92 | 6,634.62 | 1,455.19 | 78.07 |
| MATH500 | 86.40 | 95.20 | 8,419.77 | 2,179.18 | 74.12 |
| GPQA | 55.05 | 68.69 | 11,911.22 | 5,888.60 | 50.56 |
| MedQA | 78.33 | 86.00 | 4,822.44 | 1,609.66 | 66.62 |
| HumanEvalPlus | 57.93 | 90.24 | 17,765.03 | 1,751.82 | 90.14 |
| MBPPPlus | 60.85 | 77.51 | 9,802.72 | 1,907.96 | 80.54 |
For example, on HumanEvalPlus, accuracy increases from 57.93% to 90.24% while generation length decreases by 90.14%. Improvements also extend to larger benchmarks such as GSM8K, MATH500, GPQA, and MedQA. Since Soft Thinking already incorporates Cold Stop, these results show the practical value of E-TRACE's event-level intervention beyond the existing low-entropy stopping baseline. The Phi-4-Reasoning results provide additional evidence of the accuracy–efficiency benefit; the explicit paired recovery counts above are for Qwen3-8B.
Response to Reviewer rgHN (3)
Q3 & W3 Defination of the reasoning length (#L)
Thank you for pointing that out. The reported #L includes all executed generation steps, including the looping phase discarded by rollback. It measures the average total physical generation length, covering both the thinking and answer-generation phases:
\text{\\#}L = \frac{1}{N} \sum_{i=1}^{N} L_i^{\mathrm{phys}}, \qquad L_i^{\mathrm{phys}} = T_i + A_i, \qquad L_i^{\mathrm{logical}} = T_i - D_i + A_i.
As distinguished in Section 6.3, logical length measures the trajectory retained after rollback. If steps are discarded, then
For illustration, suppose a run executes thinking steps in the thinking phase, then rolls back by (or discards) 300 steps because of loop detection, and then generates a 200-token answer. Its reported physical length is 1,200, while its logical length is 900. The 300 discarded steps are fully charged to #L.
The reported length reductions therefore reflect fewer executed generation steps across the complete run, with all pre-intervention generation steps still counted. E-TRACE can avoid further unproductive thinking by intervening earlier, and its intervention can also change the subsequent answer length. Removing an already generated suffix does not itself reduce the physical count.
Response to Reviewer rgHN (1)
Q1&W1: Detailed information on the loop
Thank you for asking us to quantify the prevalence of this issue. We analyzed Qwen3-8B on all eight evaluated benchmarks. For soft-token reasoning, we use uninterrupted rollouts to measure native termination behavior. We define no-exit as exhausting the maximum budget of the model (with a 32k-step limit) without producing the native
</think>token.The soft-token results show that no-exit occurs in a subset of examples across all eight benchmarks and all three task families, with rates ranging from 1.29% to 24.24%. Its prevalence varies substantially across benchmarks: for example, GPQA has a higher observed rate than GSM8K. Soft-token reasoning also exhibits higher no-exit rates than either discrete baseline.
To examine whether these failures to exit are accompanied by repetitive loops, we computed three repetition diagnostics on all no-exit trajectories of each method. Discrete CoT uses the generated token sequence; soft-token reasoning uses the per-step top-1 token as the sequence. For a thinking sequence of length T, let Nn=T−n+1 be the number of n-gram occurrences, Un the number of distinct n-grams, and Rn=Nn−Un the number of occurrences beyond their first occurrence:
SER@L (suffix extension rate) measures how often a step continues a previously observed suffix of at least L tokens with the same next token. Under our convention, SER@L=RL+1/T; higher values indicate more steps continuing repeated sequences.
rep-n measures the fraction of repeated n-gram occurrences: rep-n=Rn/Nn. Higher values indicate greater repetition of token blocks.
D-n (Distinct-n) measures n-gram diversity: Dn=Un/Nn=1−rep-n. Lower values indicate less varied content.
We count overlapping n-grams within each generated thinking sequence, excluding the prompt, and average each metric over that method's no-exit samples respectively.
It can be seen that soft-token no-exit trajectories exhibit substantially stronger repetition than the discrete no-exit subsets. Thus, much of the soft-token thinking sequence repeatedly extends existing content. Together with the paper's event-level trajectory analysis, these results support persistent loop lock-in as a mechanism underlying failure to exit, and show more intense repetition in the observed soft-token failure cases. We report the statistics of E-TRACE to solve the problems in the next response.