- Rating
- 3(+1)
- Confidence
- 3
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 1
Read review
Paper summary
The paper addresses inference-time control of a pretrained sampler. Given a sampler with output law p₀ and a non-negative reward or likelihood G₀, the goal is to sample from G₀p₀ without retraining. Specifically, ENCORE has each replica carry its partial denoising trajectory. Moving a replica up the ladder is then a truncation, which is exactly a draw from the true time reversal. The intractable marginals cancel, and the Metropolis–Hastings ratio reduces to the SMC backward weight evaluated at two points. In the unguided case this is a ratio of potentials only. The cost is O(N²d) memory.
Critical strengths
-
Removes a real bias in the closest prior method. The construction is simple and checkable, and correctness does not depend on the model exactly reversing a noising process.
-
Broader applicability. The unguided acceptance uses only potential values. It therefore applies to distilled or deterministic samplers with no noising kernel.
Critical weaknesses
-
Limited algorithmic novelty. Path-space lifting is standard in SMC. Appendix C itself notes that the exchange weight is the SMC incremental weight on the trajectory ladder. The contribution is transplanting this into replica exchange.
-
Theory–experiment mismatch. The favourable result (finite Λ) concerns the guided variant. However, every experiment whose configuration is stated uses unguided proposals.
-
No demonstrated advantage over naive reweighting. Plain self-normalised importance sampling (SNIS) beats ENCORE in every column of Table 4 and in the fixed-budget tables. On SDXL-Turbo, best-of-64 nearly matches ImageReward (1.651 vs 1.711) with higher diversity and 100% unique outputs, versus 84% for ENCORE.
Questions for authors
-
Your main positive result, Thm. 3.2(b), shows that the guided variant has a finite communication barrier. As far as I can tell from App. E, however, all experiments use unguided proposals. Could you report results for guided ENCORE on at least one task?
-
Storage complexity and scalability to larger data. ENCORE stores the full partial trajectory at every level, so each ladder holds O(N²d) memory. The experiments keep this small: 5 levels for images, and 22–63-atom molecules elsewhere. I am concerned that the method becomes hard to apply to larger state spaces such as high-resolution images, proteins, etc.
-
Sampling efficiency. Could you report restart rates, round-trip rates, or effective sample size per NFE alongside accuracy? This would show how much of ENCORE’s output is effectively independent compared with SMC baselines.
4 comments
Response to Reviewer 45EX [2/3]
Memory (Q2)
ENCORE stores the full partial trajectory at every level, so each ladder holds memory.
We agree that ENCORE incurs an additional storage cost, as noted in our limitations (Sec. 5). For replica levels and -dimensional states, storing the paths requires sampler-state memory, compared with for CREPE and for SMC methods such as TDS with parallel particles. In exchange, the stored paths let ENCORE move up the ladder by truncation, without additional denoiser evaluations.
ENCORE can be preferable when memory or parallelism is fixed and samples are needed online. Its path storage stays constant as samples accumulate. Counting states at replica levels, a single ENCORE ladder stores fewer states than parallel SMC particles when . In our molecular tempering experiments, and each run retains samples, so .
Sampling Efficiency (Q3)
Could you report restart rates, round-trip rates, or effective sample size per NFE alongside accuracy?
To assess practical sampling efficiency, we compare unguided (U) and guided (G) ENCORE on the same 15-replica ladder. Each run used one NVIDIA H100 and retained 49,000 outputs after 1,000 warm-up half-sweeps. Tabs. 4–5 below report communication rates and effective sample size (ESS) per denoiser evaluation, as mean ± sample standard deviation over five seeds per configuration.
Table 4. Exchange rejection and communication on the same replica ladder.
| Target | Summed rejection U ↓ | Summed rejection G ↓ | Restarts/1,000 half-sweeps U ↑ | Restarts/1,000 half-sweeps G ↑ | Round trips/h U ↑ | Round trips/h G ↑ |
|---|---|---|---|---|---|---|
| ALDP 300 K | 6.185 ± 0.013 | 2.282 ± 0.089 | 11.66 ± 0.42 | 88.52 ± 2.90 | 67.66 ± 2.43 | 314.78 ± 10.08 |
| A4 500 K | 4.289 ± 0.003 | 1.629 ± 0.008 | 34.23 ± 0.49 | 114.11 ± 1.73 | 192.28 ± 5.00 | 387.13 ± 12.14 |
| A6 600 K | 3.322 ± 0.017 | 1.373 ± 0.008 | 55.21 ± 1.25 | 135.99 ± 0.52 | 298.92 ± 5.43 | 452.33 ± 3.03 |
Guidance reduces summed rejection and increases both restart frequency and round trips per hour on all three targets.
Table 5. Estimated ESS per million denoiser evaluations. One NFE is one denoiser forward evaluation on a single molecular state.
| Target | Observable | ENCORE-U ESS/million NFEs ↑ | ENCORE-G ESS/million NFEs ↑ | Paired G/U ratio ↑ |
|---|---|---|---|---|
| ALDP 300 K | Energy | 67.1 ± 20.7 | 1,136.0 ± 96.4 | 18.48× ± 6.71 |
| ALDP 300 K | Torsion | 66.4 ± 44.2 | 575.2 ± 105.8 | 13.03× ± 9.55 |
| A4 500 K | Energy | 187.5 ± 54.9 | 804.2 ± 141.1 | 4.63× ± 1.72 |
| A4 500 K | Torsion | 147.0 ± 14.2 | 414.0 ± 55.1 | 2.86× ± 0.59 |
| A6 600 K | Energy | 301.8 ± 65.7 | 999.9 ± 237.9 | 3.37× ± 0.73 |
| A6 600 K | Torsion | 237.2 ± 78.9 | 520.3 ± 85.8 | 2.46× ± 1.07 |
We estimate ESS from autocorrelations of each chronological post-warm-up trace, retaining repeated states. Torsion denotes the sine of the first backbone angle. We compute G/U ratios within each seed before averaging.
Across all three targets, guided ENCORE completed more round trips per hour and achieved lower mean error on all four accuracy metrics at matched NFE budgets (see Tab. 1 in our response to Reviewer n25t). Guidance also increased ESS per NFE for both observables in every seed on every target, with mean paired gains of 3.37–18.48× for energy and 2.46–13.03× for torsion.
Response to Reviewer 45EX [3/3]
Connection between SMC and RE (W1)
Path-space lifting is standard in SMC.
Appendix C itself notes that the exchange weight is the SMC incremental weight on the trajectory ladder. The contribution is transplanting this into replica exchange.
Path-space lifting is common in SMC, but as far as we know, it has not been used in diffusion and replica exchange before this work. SMC moves in one direction, so path space fits naturally. Replica exchange moves in both directions and is usually used to get marginal samples, not paths. Moving up the ladder needs the sampler's time reversal, which is not practical, so CREPE is only approximate. By storing the trajectory, this move becomes a truncation. This gives the first exact replica-exchange method for inference-time control, together with the communication barrier theory in Thm. 3.2.
Weight matching in SMC is expected, as it is in parallel tempering and annealed SMC. The two approaches organise the same computation differently: SMC runs particles in parallel across sequential levels, while replica exchange runs levels in parallel across sequential iterations. We are not claiming to outperform SMC. We provide the exact counterpart on the replica-exchange side, which was not available before.
SNIS and BoN (W3)
Plain self-normalised importance sampling (SNIS) beats ENCORE in every column of Table 4 and in the fixed-budget tables.
We agree and should have stated this more clearly. SNIS works well when the proposal covers the target, but coverage can deteriorate under stronger tilts or in higher dimensions.
To test this in practice, we compared SNIS and ENCORE on eight tasks: a 1D Gaussian shifted by , three peptide tempering tasks, and four BioEmu protein benchmarks biased towards lower native-contact fractions (see our response to Q2 above). Guided ENCORE had lower mean error than SNIS on all eight (see Tab. 1 above).
On SDXL-Turbo, best-of-64 nearly matches ImageReward (1.651 vs 1.711) with higher diversity and 100% unique outputs, versus 84% for ENCORE.
We agree that Best-of-N (BoN) offers a strong reward–diversity trade-off on the image tasks. For each prompt, BoN returns the highest-reward image from each of 32 independent groups of 64 draws, so its outputs are independent and distinct. ENCORE instead records two outputs from each of 16 independent chains; the two coincide when the chain's target-level replica has not changed, lowering uniqueness and diversity scores. At a fixed budget, fewer but longer chains trade diversity for reward: with 8 chains instead of 16, mean ImageReward rises from 1.711 to 1.745, while DINO-Vendi falls from 3.02 to 2.78 (Tab. 13 of the updated manuscript).
Unlike BoN, ENCORE and the Monte Carlo baselines target . However, these metrics cannot assess target fidelity for any method: reward and generic diversity do not measure discrepancy from , and no reliable target reference is available. We therefore interpret the image results as showing that ENCORE is competitive and works with distilled samplers, including DMD2, whose native sampler lacks the noising kernel CREPE requires. We will clarify this scope in the manuscript.
Thank you again for your review. We would be happy to clarify anything further during the discussion period.
Thanks for the addtional experiments and clarification from authors. I believe include the new experiments, clarifications about memory usage, and further performance analysis in the manuscript will make this paper better. I will increase my score.
Response to Reviewer 45EX [1/3]
We thank the reviewer for their time and feedback on our manuscript. We address each comment below, providing clarification and additional results where appropriate.
Guided ENCORE (Q1, W2)
Yes. Tab. 1 below compares guided and unguided ENCORE on eight tasks. Guidance lowered mean error on seven of the eight. In molecular tempering, guided ENCORE achieved lower mean error on all four metrics for all three molecules, under the same physical-energy target and a budget of 4.22M denoiser evaluations per run. Please see Tab. 1 in our response to Reviewer n25t for these results alongside CREPE, TDS and FKD.
Table 1. Sampling accuracy of SNIS and ENCORE.
Entries are mean ± standard deviation over 32 seeds (Gaussian shift), five (molecular tempering) or three (proteins).
Scalability to Larger Data (Q2)
To show that ENCORE can be applied to larger molecular state spaces, we extend our experiments to four proteins with 48–155 residues and perform enhanced sampling targeting the biased distribution
π(x)∝p0(x)exp[−α FNC(x)],
where p0 is the distribution over protein backbone conformations generated by BioEmu v1.1 using an Euler-discretised sampler, and FNC(x) is a differentiable fraction of native contacts (FNC) computed from Cα distances. We set the bias strength to α=8 for all four proteins, favouring conformations with lower FNC.
Each residue contributes three translational and three rotational degrees of freedom, giving nominal sampler-state dimensions of 288, 534, 564 and 930 for 1ENH, 1BTA, 1TTG and 2RN2. With three Cartesian coordinates per atom in the alanine-peptide experiments, the largest protein has about five times the state dimension of our largest peptide system.
Tabs. 2 and 3 below report RMSD-distribution error and backbone geometric validity as mean ± standard deviation over three seeds.
Table 2. RMSD-distribution Wasserstein error (10−3 nm; lower is better).
Table 3. Backbone geometric validity (%; higher is better).
ENCORE performs competitively across the four proteins. ENCORE-G ranks first or second in RMSD-distribution error on every protein, and ENCORE-U ranks first on the largest protein (2RN2) and third on the other three. Both variants maintain high backbone geometric validity and outperform CREPE-G and CREPE-U on both metrics for every protein.