{"id":"b2999015-5de5-4750-b19d-212ac851d3d7","arxiv_id":"2604.19635","paper_version":2,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":6.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":0,"one_line_summary":"Chunk-wise interleaved splicing of an autoregressive model enables stable streaming target speaker extraction that matches or beats offline baselines at RTF 0.248.","lead":"StarTSE adapts autoregressive generative models for real-time target speaker extraction using chunk-wise interleaved splicing and historical context refinement. It targets low-latency streaming audio where prior generative TSE systems fail due to train-inference mismatch.","discovery_kind":"new_method","skeptic_critique":{"model":"grok-4.5","headline":"Abstract-only review leaves the load-bearing empirical claim (that CIS + historical refinement closes the AR train-inference gap for streaming TSE) uninspectable; no methods, ablations, or metrics can be verified.","rationale":"The reader's UNVERDICTED / LOW-confidence stance is the correct one for an abstract-only systems paper whose load-bearing claim is empirical. The single most load-bearing concern is exactly the one the reader named: that CIS and historical refinement are asserted to close the train-inference gap, but the abstract supplies no inspectable definition, ablation, or multi-condition evidence that they do so rather than merely perform well on Libri2Mix under the reported protocol. No stronger internal inconsistency can be diagnosed without the full text; manufacturing one would violate good-faith review. Agreement with the reader is therefore full, and the verdict should remain UNVERDICTED until the paper body, metrics tables, and preferably artifacts are available for the concrete check above. Novelty and within-subfield usefulness remain plausible if the claim holds; correctness risk stays unknown.","tokens_in":1988,"tokens_out":634,"duration_ms":6299,"concrete_test":"Obtain the full paper (and code/checkpoints if released). Re-implement or re-run the AR generative baseline under the same low-latency chunk schedule without CIS/historical refinement, then with each component ablated, on Libri2Mix (and if possible a second mixture corpus). Check whether (a) the no-CIS condition shows the claimed catastrophic degradation, (b) full CIS+refinement restores 100% stability and intelligibility matching or exceeding the offline baseline at the reported RTF, and (c) results hold under at least two random seeds and an alternate latency setting. If any of (a)–(c) fails, the headline claim weakens.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is empirical: Chunk-wise Interleaved Splicing plus historical context refinement make AR generative TSE stable and competitive under streaming latency, closing the train-inference mismatch that the abstract says causes catastrophic degradation under direct adaptation. Because only the abstract is available, none of the supporting apparatus can be checked: the precise definition of the splicing schedule and how it is applied at train vs. inference time; the formulation of historical context refinement and whether it is used only at inference or also during training; the latency/chunk-size operating points; the definition of \"100% stability\"; the intelligibility and quality metrics (and whether they are reported with error bars or multiple seeds); the offline and streaming baselines and whether they are matched for model capacity and training data; and the RTF measurement protocol. Without those, it is impossible to distinguish a genuine closing of the mismatch from a Libri2Mix-specific masking effect or from an evaluation that does not stress the boundary conditions the method claims to fix. The reader's weakest_assumption correctly flags this gap; the concern is not that the idea is implausible, but that the abstract alone cannot establish that the proposed mechanisms actually do what is claimed.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.5","summary":"The manuscript proposes StarTSE, claimed to be the first autoregressive generative models tailored for streaming Target Speaker Extraction (TSE). It introduces a Chunk-wise Interleaved Splicing Paradigm intended to enable efficient, stable streaming inference and a historical context refinement mechanism to reduce boundary discontinuities between extracted segments. On Libri2Mix, the abstract reports that an AR generative baseline degrades at low latencies, whereas the proposed approach maintains 100% stability and superior intelligibility, with streaming results comparable to or surpassing offline baselines and an RTF of 0.248 on consumer-level GPUs. The central claim is that these mechanisms close the train–streaming inference mismatch that otherwise causes catastrophic degradation under direct adaptation of AR generative TSE models.","tokens_in":2280,"tokens_out":839,"duration_ms":16495,"significance":"If the empirical claims hold under full scrutiny, the work would be a meaningful systems contribution: it would show that AR generative backbones can be made viable for latency-sensitive TSE rather than remaining offline-only, and it would introduce a concrete inference-time splicing schedule plus historical refinement as a reusable pattern for streaming generative speech separation/extraction. The reported combination of 100% stability, offline-comparable or better quality/intelligibility, and RTF 0.248 on consumer GPUs would be practically relevant for real-time applications. Those strengths, however, rest entirely on experimental evidence that is not inspectable from the abstract alone.","major_comments":[{"comment":"The central load-bearing claim—that Chunk-wise Interleaved Splicing plus historical context refinement closes the AR train–streaming mismatch enough to avoid catastrophic degradation—cannot be assessed from the abstract. A full methods section must define the splicing schedule (chunk size, interleaving order, train-time vs. inference-time application), the exact formulation of historical context refinement (inputs, update rule, whether used in training), and the latency operating points. Without that, the claim remains an untested axiom rather than a demonstrated result.","section":null},{"comment":"The abstract’s headline metrics (100% stability; superior intelligibility; streaming results comparable to or surpassing offline baselines; RTF 0.248) are not supported by inspectable tables, ablations, error bars, or seed statistics in the material provided. The definition of “100% stability,” the intelligibility/quality metrics used, the matched offline and streaming baselines (capacity, data, latency), and the RTF measurement protocol are all load-bearing for the central claim and must be present and reproducible.","section":null},{"comment":"Direct adaptation is said to cause “catastrophic inference performance degradation,” yet no quantitative characterization of that failure mode (e.g., collapse rate vs. latency, spectrogram/artifact examples, or ablation removing CIS or historical refinement) is available here. Establishing that the proposed mechanisms specifically fix that failure—rather than a Libri2Mix-specific or evaluation-protocol effect—is essential to the paper’s contribution.","section":null}],"minor_comments":[{"comment":"Abstract-only review: figure/table numbering, notation consistency, related-work placement, and writing polish cannot be checked until the full manuscript is provided.","section":null},{"comment":"When the full text is available, ensure the abstract’s coined terms (Chunk-wise Interleaved Splicing Paradigm; historical context refinement) map one-to-one onto named algorithms/equations so readers can locate the formal definitions.","section":null}],"recommendation":"uncertain","confidential_remarks":"Only the abstract was available for this review (full text marked unavailable). I cannot responsibly recommend accept/minor/major/reject on empirical systems claims without methods, results tables, ablations, and baselines. Recommendation is therefore uncertain pending the complete manuscript. If the full paper is supplied, a normal re-review against the same load-bearing points is appropriate; the idea is plausible but currently uninspectable."},"author_rebuttal":null,"desk_editor":{"model":"grok-4.5","letter":"The one thing to know: this is an abstract-only claim that the first autoregressive models tailored for streaming target speaker extraction, using a Chunk-wise Interleaved Splicing Paradigm plus historical context refinement, keep 100% stability and offline-comparable quality on Libri2Mix at RTF 0.248. That is a real systems problem—generative TSE usually needs global context and collapses under low-latency streaming—and if the numbers hold it matters for hearing-assist and live comms products.\n\nWhat looks new is the specific inference-time splicing schedule and the historical refinement step aimed at boundary discontinuities. Framing AR generative backbones as viable for latency-sensitive TSE is a clean application move, not a paradigm shift, but it is a legitimate first-in-class claim for this setting. The abstract is clear about the train-inference mismatch and does not hide the degradation of naive adaptation.\n\nThe soft spot is structural and large: we only have the abstract. None of the supporting apparatus is checkable—exact splicing schedule at train vs. inference, whether historical refinement is training-time or inference-only, chunk sizes and latency points, definition of “100% stability,” metrics with error bars or seeds, capacity-matched offline/streaming baselines, or RTF protocol. So we cannot yet tell whether CIS + refinement actually closes the mismatch or merely masks it on Libri2Mix. That is the reader’s weakest assumption, and the stress-test is right: the central claim is empirical and currently uninspectable. No circularity or free-parameter smell is visible from the abstract; it is a standard empirical systems paper.\n\nThis is for speech-separation and real-time audio people who already care about generative TSE. It is not for a general ML reading group until methods and ablations exist. I would send it to peer review rather than desk-reject: the problem is real, the proposed fix is concrete, and a serious referee can demand the missing tables and code. I would not cite it yet and would not bring the abstract alone to reading group. If the full paper ships clean ablations and artifacts, that changes.","headline":"Abstract-only systems paper claiming first AR streaming TSE via chunk-wise interleaved splicing; idea is plausible and within-subfield useful, but the load-bearing empirical claims are uninspectable.","tokens_in":2915,"tokens_out":538,"would_cite":false,"duration_ms":5573,"reading_group":"no","serious_thinker":"unclear","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.5","headline":"Autoregressive models can do streaming target speaker extraction without collapse, via chunk-wise interleaved splicing and historical refinement.","keywords":["target speaker extraction","streaming TSE","autoregressive generative models","chunk-wise interleaved splicing","historical context refinement","real-time factor","Libri2Mix","speech separation"],"falsifier":"On Libri2Mix (or a comparable mixture corpus), measure whether a naive streaming AR baseline still collapses while the interleaved-splicing model keeps near-offline SI-SDR/PESQ/STOI and 100 percent stability at the reported chunk latencies; any large residual gap or instability under those conditions would falsify the claim.","tokens_in":2879,"feed_emoji":"🎙️","tokens_out":846,"duration_ms":8304,"temperature":0.7,"pith_summary":"Generative models have raised the bar for Target Speaker Extraction (TSE), which pulls one speaker's voice out of a mixture, but they normally need the full recording and therefore cannot run in real time. Directly forcing them into streaming mode usually produces catastrophic quality loss because the model never saw partial inputs during training. This paper presents the first autoregressive models built expressly for streaming TSE. The key device is a Chunk-wise Interleaved Splicing Paradigm that feeds the model successive short chunks while still letting it generate speech autoregressively, together with a historical-context refinement step that smooths the joins between chunks. On the Libri2Mix benchmark the method keeps 100 percent stability and higher intelligibility than a naive streaming baseline; the streaming scores match or beat offline systems, and the real-time factor on ordinary GPUs is 0.248. The result shows that autoregressive generative backbones can be made practical for latency-sensitive speech applications once the training-inference mismatch is closed by the right splicing scheme.","feed_headline":"Streaming TSE works with autoregressive models via chunk splicing","feed_subtitle":"Interleaved chunks plus history keep 100% stability and match offline quality at RTF 0.248","key_machinery":"Chunk-wise Interleaved Splicing Paradigm: successive short audio chunks are interleaved and spliced so that an autoregressive generative backbone can produce continuous extracted speech while remaining compatible with its training regime; historical context refinement then re-uses prior frames to remove boundary discontinuities.","core_discovery":"The first autoregressive models tailored for streaming target speaker extraction, using Chunk-wise Interleaved Splicing plus historical context refinement, maintain 100 percent stability and superior intelligibility on Libri2Mix, with streaming results comparable to or surpassing offline baselines and an RTF of 0.248 on consumer GPUs.","pith_inferences":["The splicing-plus-refinement pattern may transfer to other autoregressive generative audio tasks that suffer the same train-inference mismatch (e.g., streaming speech enhancement or source separation).","If historical refinement is the main coherence mechanism, its computational cost and memory footprint will set a practical lower bound on usable chunk size for on-device deployment.","Empirical success on Libri2Mix leaves open whether the same stability holds for more adverse acoustic conditions (noise, reverberation, more than two speakers).","A natural next measurement is the latency-intelligibility Pareto front under varying chunk lengths to quantify how much quality is traded for each millisecond of latency."],"forward_implications":["Streaming AR-TSE can run on consumer GPUs at RTF 0.248 without the collapse previously observed for generative models.","Latency-sensitive applications (hearing aids, live conferencing, on-device voice interfaces) can adopt generative extraction quality rather than remaining limited to non-generative pipelines.","Offline-quality intelligibility and stability become achievable under streaming constraints once the splicing and refinement steps are applied.","The same paradigm supplies a concrete template for converting other full-context generative speech models into streaming versions."],"fun_headline_variants":["AR models stream TSE stably with chunk-wise interleaved splicing","First autoregressive TSE for streaming via interleaved chunk splicing","Chunk splicing plus history keeps AR TSE at 100% stability","Streaming AR TSE matches offline quality at RTF 0.248","Interleaved splicing unlocks real-time autoregressive speaker extraction"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"That the train-inference mismatch for autoregressive generative TSE is sufficiently closed by chunk-wise interleaved splicing and historical context refinement alone, so low-latency streaming does not reintroduce the catastrophic degradation the authors attribute to direct adaptation.","fun_headline_variants_meta":{"raw":{"variants":["AR models stream TSE stably with chunk-wise interleaved splicing","First autoregressive TSE for streaming via interleaved chunk splicing","Chunk splicing plus history keeps AR TSE at 100% stability","Streaming AR TSE matches offline quality at RTF 0.248","Interleaved splicing unlocks real-time autoregressive speaker extraction"]},"model":"grok-4.5","effort":"low","cost_usd":0.003552,"raw_usage":{"total_tokens":1156,"prompt_tokens":750,"num_sources_used":0,"completion_tokens":86,"cost_in_usd_ticks":35520000,"prompt_tokens_details":{"text_tokens":750,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":320,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":750,"tokens_out":86,"duration_ms":4057,"temperature":1.0,"reasoning_tokens":320,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-12T18:49:39.531550+00:00","model_set":{"reader":"grok-4.5"},"falsifier":"On Libri2Mix (or a comparable mixture corpus), measure whether a naive streaming AR baseline still collapses while the interleaved-splicing model keeps near-offline SI-SDR/PESQ/STOI and 100 percent stability at the reported chunk latencies; any large residual gap or instability under those conditions would falsify the claim.","supporting_citations":[],"review_version":2}