- Rating
- 2
- Confidence
- 2
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 2
Read review
Paper summary
This paper proposes a training-free optimization framework for latent space reasoning, Trajectory-Invariant Latent Refinement (TILR), aiming to address the sensitivity of existing latent space reasoning correction methods to hyperparameters, reference model selection, and input perturbations. Specifically, TILR first extracts a stable correction subspace from the representational differences between strong and weak reference models using Singular Value Decomposition (SVD), and then dynamically adjusts the correction strength of the latent states using an Adaptive Stability Gate, thereby improving the consistency and robustness of the reasoning trajectory. The authors conducted experiments on six mathematical and symbolic reasoning benchmarks and further validated the applicability of the method on different model architectures. Experimental results show that TILR achieves a certain accuracy improvement compared to existing methods while maintaining low additional computational overhead, and significantly improves answer consistency and trajectory stability under input perturbations.
Critical strengths
- Training-free optimization has practical value. TILR can improve latent space inference without updating model parameters or performing backpropagation, offering certain deployment convenience.
- The adaptive correction mechanism is value. The Adaptive Stability Gate adjusts the update intensity based on the alignment between the correction direction and the learning subspace, helping to suppress unreliable corrections under low alignment conditions.
- The paper has a complete and clear structure.
Critical weaknesses
- The theoretical basis for invariant subspaces is insufficient. Current cannot directly guarantee their invariance to the input or reference model.
- The reliability of adaptive gating is insufficient. Subspace alignment does not necessarily represent the correctness or effectiveness of the correction direction, and gating offers only a small accuracy gain compared to individual projection, lacking a more direct mechanism for verification.
- On some datasets, TILR shows only a small improvement over Refinement.
- The experimental model scale and task coverage are limited. The main experiments are primarily based on GPT-2, with the appendix only extending to 1B–1.5B models. Furthermore, the tasks focus on mathematical and symbolic reasoning, which is insufficient to verify the method's generalization ability on larger-scale models and complex reasoning tasks.
Questions for authors
- It is recommended to analyze the correlation between subspace alignment and actual correction gains.
- Further validate the effectiveness of TILR on 7B or larger models, and on more complex tasks such as multi-hop inference and open-domain inference?
- It is recommended that the authors report the complete inference cost, including forward computation of the reference model and subspace calibration, and provide an end-to-end efficiency comparison with the original Coconut.
4 comments
Follow-up on Our Response and Additional Results
Dear Reviewer,
Thank you again for your review. We have conducted the additional experiments suggested in your review and incorporated the results into the revised manuscript.
We would appreciate it if you could revisit our response when you have a chance. We would be happy to address any remaining concerns during the discussion period.
I appreciate the authors' responses and supplementary experiments, which have clarified some of the questions. However, the current manuscript still has some shortcomings. Existing evidence supports effective filtering under specific perturbations, but its robustness and transferability are limited. Furthermore, the current manuscript's analysis of the method's applicability conditions and failure scenarios is insufficient. Additionally, based on the presented experimental results, the accuracy improvement on larger-scale models is limited, while the complete inference process still incurs significant additional computational overhead. Considering the above issues and the comments of other reviewers, I maintain the original score.
Clarification on Remaining Concerns
Dear Reviewer,
Thank you for revisiting our response and additional experiments. We appreciate your comments. We would like to briefly clarify the remaining points, as the revised manuscript and rebuttal experiments directly evaluate and address each of these concerns.
| Remaining concerns | Evidence and results directly addressing the concerns |
|---|---|
| Robustness / transferability remains limited | The added experiments tests transfer across 7 alternative references, including an independent training seed (fixed : 27.63% vs. 27.85% with per-reference recalibration); across benchmarks without recalibration (SVAMP: +1.7 pp, ; GSM-Plus: +0.8 pp, ); across GPT-2, Llama, and Qwen architectures up to 7B; and across latent-reasoning frameworks. On CODI, TILR reduces trajectory variance by 47-48% while retaining/improving accuracy. These results show that TILR's robustness effect transfers successfully across references, training seeds, benchmarks, architectures, model scale, and latent-reasoning frameworks. |
| Accuracy improvement at larger scale is limited | The paper's objective is robustness while preserving accuracy, rather than maximizing accuracy gains. At 7B, the strongest reference provides only 0.7 pp validation headroom; nevertheless, TILR preserves accuracy (33.51% 33.66%) while reducing paraphrase trajectory variance by 11%. Thus, the 7B result directly validates the central robustness claim at larger scale. |
| Complete inference incurs significant overhead | The end-to-end experiment directly measures this concern, including all reference forward passes. TILR requires 68.7 ms/question vs. 67.9 ms/question for unconstrained Refinement, i.e., only 1.2% incremental latency. Both methods use the same 24.3 forward passes; therefore, the substantial cost relative to Coconut (28.8 ms) comes from the contrastive reference inference already required by the baseline Refinement method, not from TILR's subspace projection or adaptive gating. TILR is also 28.5% faster than CoT in the same measurement, and calibration is a one-time ~10 s cost. Thus, the revised end-to-end measurement directly quantifies the computational concern and shows that the proposed TILR mechanism itself adds negligible overhead while providing significant robustness improvement. |
| Applicability conditions / failure scenarios are insufficiently analyzed | We directly experiment the intended failure regime by progressively degrading reference reliability. As the reference deteriorates, the gate responds as designed () and removes 60-84% of projection-induced degradation. Under matched correction magnitude, reference noise causes only a 4.8 pp accuracy loss with gating versus 12.7 pp without, ruling out smaller updates as the explanation. This experimentally characterizes when gating becomes important and how TILR behaves when its correction signal becomes unreliable. The revised Limitations separately scopes conditions not established by our experiments, including larger distribution shifts, different reasoning domains, calibration dependence, and the distinction between invariance and correctness. |
We agree that broader evaluation across every possible model, domain, and distribution shift would provide additional evidence. However, these settings are outside the paper's stated scope rather than assumptions required by its claims. Within the stated scope, the revised manuscript now directly addresses the robustness, transferability, scale, computational cost, and failure-regime questions raised above.
If there is a specific stated claim that you believe remains unsupported after these experiments, we would greatly appreciate clarification so that we can address it precisely during the remaining discussion period. Thank you.
Response to Reviewer niJG: Experimental Validation of Invariance, Gate Reliability, and Scaling to 7B
Thank you for the constructive feedback and recognition of our work. We have addressed your concerns with the experimental results below.
W1. Theoretical basis for invariance. Appendix B makes no formal claim that Sr contains causally invariant features. Eq. 12 instead gives a conditional result: when two corrections differ only outside Sr and have equal norm, TILR applies the same projected correction, introducing no additional divergence beyond the backbone. Appendix D bounds step-size sensitivity by γ2 of unconstrained refinement. The empirical question is how much real variation lies outside Sr. We now measure this on 920 paraphrased GSM8K questions (lower filters more):
W2. Reliability of the gate. γ measures geometric support for the current reference signal. We directly test whether it tracks reference reliability by progressively corrupting the reference and comparing projection-only refinement with TILR at the same η (GSM8K test, n = 1,319, change vs Coconut in points):
Under noise γ falls to the value of a direction with no preference for the subspace (√(171/768) = 0.47), and the step halves. In a second run where gated and ungated corrections are matched in average step size, 3× noise costs 4.8 points with the gate and 12.7 without. The gate's gains are in robustness, not average accuracy: the hyperparameter sensitivity gap is 10–18× smaller (Table 4).
W3. Small improvement over Refinement. We agree that the mean-accuracy improvement over Refinement is modest; this is consistent with the paper's stated objective of improving robustness while preserving accuracy, and its margin is largest there:
W4 / Q2. Scale and tasks. Appendix H reports Qwen2.5-Math-1.5B and Llama-3.2-1B, with gains in accuracy, M3 and M4 on all six backbone-benchmark pairs. New: we trained Coconut on Qwen2.5-7B and ran the full pipeline. TILR gives comaparable or slightly better accuracy (GSM8K 33.66% vs 33.51%; SVAMP 69.9% vs 69.7%) while reducing trajectory variance under paraphrasing by 11% (243.8 vs 272.8). For multi-hop inference, ProsQA and ProntoQA are multi-hop deductive tasks, where TILR is the best latent refinement method (Table 1: 90.4%, 94.7%).
Q1. Alignment vs correction gain. Across controlled reference corruptions, lower γ coincides with lower correction reliability. Crucially, this is not explained merely by smaller steps: when gated and ungated updates are matched in average step magnitude, 3× reference noise reduces accuracy by 4.8 points with gating versus 12.7 without.
Q3. End-to-end cost. One RTX 6000 Ada, batch size 1, GSM8K test (200 questions), all reference forward passes included:
The projection and gate add 1% over refinement; TILR is 28.5% faster than CoT. Calibration takes approximately 10 seconds.
In summary, we provide direct evidence addressing each concern:
We hope these additional results address all your concerns and clarify that TILR's contribution is robust refinement rather than mean-accuracy improvement.