{"id":"6eb346d7-5de8-490c-9c65-2e6f8da2f4af","arxiv_id":"2606.22239","paper_version":1,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":7.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":0,"one_line_summary":"Derives an interacting-particle diffusion sampler via Doob h-transform that targets batches with high empirical variance under a linear feature map.","lead":"The paper introduces a variance-weighted batch distribution for diffusion models that favors diverse samples after a linear feature map, derived via Doob h-transform as an interacting particle system with repulsion and curvature terms. A smart generalist might read it for a principled alternative to heuristic diversity methods in generative modeling applications like design or optimization.","discovery_kind":"new_method","skeptic_critique":{"model":"grok-4.3","headline":"Closed-form derivation of Doob h-transform for variance-weighted batch target may embed unstated restrictions on the linear feature map or diffusion process.","rationale":"The reader's weakest_assumption isolates exactly the step whose validity determines whether the sampler is a true Doob transform rather than a heuristic. Verifying the derivation on a tractable case directly tests whether that step is secure; agreement is therefore high and the verdict moves from UNVERDICTED to CONDITIONAL pending the check.","tokens_in":1608,"tokens_out":327,"duration_ms":12682,"concrete_test":"Take the explicit h-transform formula claimed in the paper (likely Eq. (X) in the methods), substitute a simple linear Φ (identity) and Ornstein-Uhlenbeck diffusion with known transition densities, and recompute the resulting drift correction analytically; if the repulsion term fails to match the claimed interaction or requires truncation, the closed-form property does not hold generally.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central claim requires that the target batch distribution (product of independent marginals reweighted by empirical variance after a fixed linear map Φ) admits an explicit Doob h-function whose gradient yields only the stated repulsion-plus-curvature correction. This holds only if the h-transform can be obtained exactly from the Fokker-Planck or backward Kolmogorov equation without series expansions, data-dependent approximations, or implicit assumptions that Φ commutes with the diffusion operator. The abstract asserts a compact closed form, but any deviation from linearity of Φ or from the precise variance functional would invalidate the interaction term.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.3","summary":"The manuscript proposes variance-tilted diffusion models to generate diverse batches from diffusion models. It defines an explicit target batch distribution that reweights independent marginals by the empirical variance of a prescribed linear feature map Φ, then derives the corresponding sampler as the Doob h-transform of the independent diffusion dynamics. The resulting correction consists of an interaction term that repels posterior denoised means and a curvature term that shifts particles toward regions of higher feature variance, yielding an interacting-particle system with a transparent probabilistic target.","tokens_in":1741,"tokens_out":539,"duration_ms":19056,"significance":"If the closed-form derivation of the h-transform holds exactly, the work supplies a principled, non-heuristic alternative to repulsive-drift methods for diversity in generative sampling. The explicit target distribution and compact correction terms would constitute a clear advance for applications that require controlled batch diversity (e.g., molecular design, prompt variation). The absence of free parameters in the target specification is a notable strength.","major_comments":[{"comment":"§3 (Doob h-transform derivation): the claim that the h-function admits an exact closed-form gradient yielding only the stated repulsion-plus-curvature correction must be verified against the Fokker-Planck or backward Kolmogorov equation for the variance-weighted product measure. Any implicit assumption that Φ commutes with the diffusion operator or that the variance functional remains linear under the noising process would invalidate the compact form; the manuscript should exhibit the explicit steps without series expansions or data-dependent fitting.","section":"§3"},{"comment":"Definition of the target (likely Eq. (target) or §2.2): the variance-weighted batch distribution is specified via an empirical variance after a fixed linear map Φ. The manuscript must confirm that this target remains a valid probability measure for arbitrary batch size and that the corresponding h-transform reduces exactly to the claimed interaction term without additional normalization constants that depend on the data.","section":"§2.2"}],"minor_comments":[{"comment":"Notation for the posterior denoised means and the curvature term should be introduced with explicit dependence on the diffusion time t to avoid ambiguity when the correction is applied at different noise levels.","section":null},{"comment":"The abstract states the sampler is 'derived as the Doob h-transform'; the introduction or §1 should include a one-sentence pointer to the precise location of the full derivation for readers who wish to verify the compact form immediately.","section":null}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":{"model":"grok-4.3","summary":"We thank the referee for the careful reading and positive assessment of the significance of the work. We address each major comment below, providing the requested verification and clarifications.","responses":[{"response":"The h-transform derivation in §3 follows directly from the Doob formula applied to the product diffusion on the batch under the variance-weighted target. Because Φ is linear, the noising process on the transformed features Φ(x_t) is an Ornstein-Uhlenbeck process with the same variance schedule, so the empirical variance functional remains exactly quadratic in the batch means at every time. We verify the gradient of log h by substituting the weighted measure into the backward Kolmogorov equation and differentiating: the first-order term produces the repulsion between posterior means, while the second-order term produces the curvature correction. The algebra is exact (no expansions or fitting) and is now exhibited in full in the revised §3 together with a new appendix containing the Fokker-Planck verification.","revision_made":"yes","referee_comment":"[§3] §3 (Doob h-transform derivation): the claim that the h-function admits an exact closed-form gradient yielding only the stated repulsion-plus-curvature correction must be verified against the Fokker-Planck or backward Kolmogorov equation for the variance-weighted product measure. Any implicit assumption that Φ commutes with the diffusion operator or that the variance functional remains linear under the noising process would invalidate the compact form; the manuscript should exhibit the explicit steps without series expansions or data-dependent fitting."},{"response":"The target is p(batch) ∝ [∏_i p(x^i)] · Var_Φ({Φ(x^i)}), normalized by the constant Z = E[Var_Φ] taken over independent draws from p. This is a valid probability measure for any finite B ≥ 2 because Var_Φ ≥ 0 and Z is finite and positive under standard moment assumptions on p. In the h-transform the constant Z cancels identically in the ratio that defines h, so the resulting SDE contains no data-dependent normalization factors; the interaction term arises solely from the quadratic structure of Var_Φ. A clarifying sentence confirming these facts has been added to §2.2.","revision_made":"partial","referee_comment":"[§2.2] Definition of the target (likely Eq. (target) or §2.2): the variance-weighted batch distribution is specified via an empirical variance after a fixed linear map Φ. The manuscript must confirm that this target remains a valid probability measure for arbitrary batch size and that the corresponding h-transform reduces exactly to the claimed interaction term without additional normalization constants that depend on the data."}],"tokens_in":1282,"tokens_out":566,"duration_ms":31365,"standing_objections":[]},"desk_editor":{"model":"grok-4.3","letter":"The one or two things to know: the authors construct a variance-tilted target distribution for batches of diffusion samples and derive the corresponding Doob h-transform of the independent dynamics, resulting in a repulsion term on the posterior denoised means plus a curvature term for higher variance regions.\n\nThis construction appears new, as it uses the h-transform on batch statistics rather than standard approaches. It does well in giving an explicit target that is not data-dependent in a fitted sense, which keeps the circularity burden low. The probabilistic framing is a step up from heuristic repulsive drifts.\n\nWhat the paper does well is connecting the diversity goal directly to the h-transform machinery and extracting interpretable correction terms. If the derivation is solid, it offers a principled tool for generating diverse candidates.\n\nThe soft spots center on the closed-form claim. The abstract states the correction has a compact form, but the stress-test note correctly flags that this may embed restrictions, like the linear map commuting with the diffusion operator or the variance functional behaving exactly as assumed. Without the full steps, it's difficult to confirm there are no hidden approximations or series expansions. The soundness is moderate because of this.\n\nThe work is for specialists in generative ML sampling. It would interest readers working on particle methods or Doob transforms in stochastic differential equations.\n\nThe paper shows clear thinking on its own terms, so it deserves peer review.\n\nRecommendation: Yes, it should go to referees for a full check of the math.","headline":"The paper derives a Doob h-transform for a variance-weighted batch target in diffusion models to produce an interacting sampler with repulsion and curvature terms, but the exactness of that compact form needs verification against possible restrictions on the feature map.","tokens_in":2212,"tokens_out":393,"would_cite":false,"duration_ms":29482,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.3","headline":"Diffusion models obtain diverse batches by tilting the target to high empirical variance and deriving the sampler as the corresponding Doob h-transform.","keywords":["diffusion models","diverse sampling","interacting particles","Doob h-transform","variance-weighted distribution","batch generation","generative models"],"falsifier":"Simulate both the variance-tilted interacting sampler and independent sampling on a low-dimensional multimodal target, then measure whether the empirical variance of generated batches under the tilted sampler exceeds that of independent sampling by the amount predicted by the explicit target measure.","tokens_in":2519,"feed_emoji":"","tokens_out":657,"duration_ms":16807,"temperature":0.7,"pith_summary":"The paper establishes that independent diffusion trajectories can be corrected into an interacting particle system whose joint law targets collections of samples with large spread after a fixed linear feature map. The correction consists of an explicit repulsion between posterior denoised means plus a curvature term that shifts particles toward regions of higher feature variance. A reader would care because many generation tasks require spread-out candidates rather than independent draws that may repeat similar modes. The construction keeps the target distribution fully specified and derives the dynamics in closed form without data-dependent fitting or heuristics. This replaces ad-hoc repulsion with a transparent probabilistic objective on batches.","feed_headline":"Diffusion sampling gains explicit diversity through variance tilt","feed_subtitle":"A closed-form Doob h-transform adds repulsion between denoised means and curvature toward high-variance regions so batches spread rather tha","key_machinery":"The Doob h-transform of the independent diffusion process with respect to the variance-weighted batch distribution, which supplies the closed-form interaction and curvature correction terms.","core_discovery":"The central claim is that the Doob h-transform of independent diffusion dynamics with respect to an explicitly specified variance-weighted batch distribution produces a compact correction: an interaction term that repels posterior denoised means together with a curvature term that moves particles toward higher feature variance, thereby yielding an interacting-particle sampler whose stationary law is the desired variance-tilted batch measure rather than a heuristic repulsive drift.","pith_inferences":["The same construction could be applied to other explicitly specified batch objectives, such as minimum pairwise distance or coverage of a reference set.","In downstream pipelines the method might reduce reliance on post-processing steps that enforce diversity after independent generation.","Because the feature map is required to be linear, extensions to nonlinear embeddings would need additional approximation steps not derived in the paper."],"forward_implications":["Batches produced by the sampler exhibit larger spread in the chosen feature space than independent trajectories.","The joint law on collections of samples remains a well-defined probability measure at every step.","The dynamics follow directly from the chosen target without requiring separate repulsive heuristics.","All correction terms are expressed in terms of posterior means and feature variances and remain computable during sampling."],"fun_headline_variants":["Variance tilt adds repulsion to diffusion sampling","Doob h-transform targets variance in diffusion batches","Interacting diffusion corrects for higher feature variance","Curvature moves particles to high variance regions"],"cache_read_input_tokens":2112,"weakest_assumption_plain":"The variance-weighted batch distribution after the prescribed linear feature map can be written explicitly and the corresponding Doob h-transform derived in closed form without approximations.","fun_headline_variants_meta":{"raw":{"variants":["Variance tilt adds repulsion to diffusion sampling","Doob h-transform targets variance in diffusion batches","Interacting diffusion corrects for higher feature variance","Curvature moves particles to high variance regions"]},"model":"grok-4.3","cost_usd":0.004556,"raw_usage":{"total_tokens":2202,"prompt_tokens":544,"num_sources_used":0,"completion_tokens":53,"cost_in_usd_ticks":45562000,"prompt_tokens_details":{"text_tokens":544,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":1605,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":544,"tokens_out":53,"duration_ms":13632,"temperature":1.0,"reasoning_tokens":1605,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-06-26T10:46:46.111836+00:00","model_set":{"reader":"grok-4.3"},"falsifier":"Simulate both the variance-tilted interacting sampler and independent sampling on a low-dimensional multimodal target, then measure whether the empirical variance of generated batches under the tilted sampler exceeds that of independent sampling by the amount predicted by the explicit target measure.","supporting_citations":[],"review_version":1}