{"id":"4295858d-8ebd-4316-b9e8-7b483e155c20","arxiv_id":"2607.23204","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"Context-aware prefaces generated before a user finishes speaking reduce the delay to the main answer, but make the first response slightly later than a fixed filler.","lead":"A dialogue robot that starts speaking a context-aware preface before the user finishes can get the real answer out sooner, but the very first word is slightly slower than with a fixed filler. A mall field test shows this timing trade-off.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Claim that contextual prefaces speed up main response is unverified: paper omits initial-response duration, so shorter gap may not translate to earlier main-response onset.","rationale":"The reader identified the core weakness: the initial-to-main gap may be affected by the duration of the initial response. I agree, and I sharpen it: the reader's own derived 'total time to main' (1.58 vs 1.68 s) is actually L+G, which omits D. Therefore even the trade-off the reader calculates is incomplete. The paper's central claim is that contextual prefaces reduce the delay before the substantive answer, but the only evidence is a shorter gap. Since M = L + D + G, an unreported D could reverse the conclusion. This is not a matter of statistical nuance but of the primary outcome variable. The paper should report M directly. The paper has strengths: a real-world field deployment, a plausible mechanism (parallel generation during preface speech), and honest limitations. But without M, the contribution is not established. The requested change is minor (report existing log timestamps) but central. Thus the current CONDITIONAL verdict is appropriate; no change needed, but the condition is important.","tokens_in":8043,"tokens_out":4647,"duration_ms":40788,"concrete_test":"Re-analyze the interaction logs: for every response in the fixed-filler and contextual-preface conditions, extract timestamps t_user_end, t_initial_start, t_initial_end, t_main_start. Compute D = t_initial_end - t_initial_start and M = t_main_start - t_user_end. Report mean/median D and M per condition and run a Mann-Whitney U test comparing M (contextual vs fixed). If M is not significantly smaller for contextual-preface, the central claim fails. Also report D as a covariate or use a mixed-effects model to account for repeated measures.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central benefit is that contextual prefaces shorten the delay before the substantive answer (Section 3.3, Discussion). The only metric supporting this is the initial-to-main gap G, defined as time from end of initial response to start of main response. The absolute time from user utterance end to main response onset is M = L + D + G, where L is initial latency and D is initial-response duration. The paper reports L and G but not D. The mean L+G is 1.58 s for contextual-preface vs 1.68 s for fixed-filler, a 0.10 s difference. Since contextual prefaces are constrained to 'no longer than 10 Japanese characters' (Section 2.1) while fixed fillers are 'yeah'/'I see', D_ctx likely exceeds D_fixed by more than 0.10 s; e.g., 10 characters at typical speech rates is ~1.2 s, a filler ~0.3 s. If so, M_ctx > M_fixed, meaning the substantive answer actually arrives later in the contextual condition, contradicting the claimed trade-off. The paper's statement that generation overlaps with preface speech is plausible, but it does not provide the one number (M) that would verify it. Without D or M, the central claim is unsupported.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a two-stage incremental response framework for LLM-based dialogue robots. An intent readiness detector triggers generation of a short contextual prefatory response when user intent becomes predictable, while a voice activity projection (VAP) model decides the precise moment to speak that preface; the main response is generated concurrently. A field experiment in a shopping mall compared three conditions (no-filler, fixed-filler, contextual-preface). Results show that both filler conditions significantly reduce initial response latency relative to no-filler, and that contextual-preface has significantly longer initial latency than fixed-filler but a significantly shorter initial-to-main gap. Subjective ratings show no significant differences. The paper interprets this as a timing trade-off: the contextual preface masks turn-onset latency and also shortens the wait for the substantive answer, at the cost of slightly later first speech.","tokens_in":8292,"tokens_out":2568,"duration_ms":24115,"significance":"If the timing effects are robust, the work offers a practical architectural pattern for reducing perceived response delay in cascaded ASR-LLM-TTS robots without relying on generic fillers. The field deployment is a real strength, as is the honest treatment of null subjective results and the explicit acknowledgment of limitations. The paper also provides a breakdown analysis of prefatory-response failures, which is valuable for future work. However, the central quantitative claim that contextual prefaces deliver the substantive answer sooner is not directly supported by the reported metrics, because the duration of the initial response is omitted. The paper's significance therefore hinges on whether the missing timing information actually confirms the claimed trade-off.","major_comments":[{"comment":"The time from user utterance end to main-response onset is M = L + D + G, where L is initial latency, D is the duration of the initial response, and G is the initial-to-main gap. The paper reports L and G but not D. Since contextual prefaces are constrained to be at most 10 Japanese characters while fixed fillers are 'yeah'/'I see' (Section 2.1), D_ctx likely exceeds D_fixed. The reported advantage in L+G is only about 0.10 s (1.58 vs 1.68 s); if D_ctx exceeds D_fixed by more than this, the substantive answer actually arrives later in the contextual-preface condition. To support the Discussion's claim that the method shortened the delay before the substantive answer (Section 4), the authors must report D and M, or at least the absolute main-response onset time conditioned on the initial response.","section":"Section 3.2, Table 1"},{"comment":"The initial-to-main gap is measured from the end of the initial response. This makes the comparison between fixed-filler and contextual-preface structurally sensitive to the duration of the initial response: a longer initial response moves the measurement window later, mechanically reducing the gap even if main-response onset timing is identical. The paper does not control for initial-response duration or restrict the gap analysis to responses with matched initial-response lengths. The authors should either analyze the absolute main-response onset time (as in the previous comment) or rerun the gap comparison including initial-response duration as a covariate/per-utterance matched analysis.","section":"Section 3.2, metric definition"},{"comment":"The Kruskal-Wallis and Mann-Whitney tests treat each response instance as an independent sample. However, the data contain multiple response instances from the same interaction (47-75 interactions, with 188-251 response instances per condition). Responses within an interaction are likely correlated (same user, same context, same system state). This pseudoreplication can inflate the effective sample size and render the reported p-values (e.g., p < .001 for the gap difference) overly optimistic. The authors should fit a mixed-effects model with interaction as a random intercept, or at least perform a sensitivity analysis using one aggregated value per interaction (e.g., per-interaction medians) and report whether the qualitative conclusions hold.","section":"Section 3.3, statistical tests"}],"minor_comments":[{"comment":"Typo: 'ano-filler condition' should be 'a no-filler condition'.","section":"Section 3.1"},{"comment":"The figure caption appears truncated or confusing: the text '('McDonald's would be nice.')' is separated across lines, and the speech-bubble labels are not fully legible in the reproduced version. Please clean up the caption and ensure all callouts in the figure appear in the text or are explained.","section":"Figure 1"},{"comment":"The 'Trigger position' metric is defined as a character-level progress ratio, but it is not described how the character counts are computed (e.g., including punctuation, hesitations). Clarify this definition.","section":"Section 3.2"},{"comment":"The sentence 'both fixed-filler and contextual-preface were significantly faster than no-filler' lacks the actual p-values for those pairwise comparisons; only the fixed-filler vs contextual-preface comparison is reported (p=.027). Please report all pairwise results or state that they were significant at the Bonferroni-corrected threshold.","section":"Section 3.3"}],"recommendation":"major_revision","confidential_remarks":"The central issue is the missing initial-response duration, which is likely available in the system logs and could be reported in a revision. If the omitted metric turns out to contradict the claimed trade-off, the manuscript's main conclusion would change. The pseudoreplication concern is also important given the large number of responses per interaction. I would encourage the editor to request this revision rather than reject, because the field experiment and system architecture are otherwise solid."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"What's actually new here is a two-stage response design: an intent-readiness detector triggers a short context-aware preface before the user finishes speaking, and a VAP model decides when to deliver it. That is paired with a field experiment in a shopping mall comparing no filler, fixed filler, and contextual preface. The experiment is real — 644 response instances across three conditions — and the authors are appropriately honest about the null subjective results.\n\nThe soft spot is load-bearing. The paper reports initial latency L and initial-to-main gap G, but not the duration D of the initial response. The time from the user's utterance end to the start of the main response is L + D + G. The paper's implicit comparison is L+G: 1.58 s for contextual preface versus 1.68 s for fixed filler. But contextual prefaces are constrained to at most 10 Japanese characters, while fixed fillers are \"yeah\" or \"I see.\" If D_ctx exceeds D_fixed by more than 0.10 s — likely, given the length constraint — then the main response actually starts later in the contextual condition. The Section 2.1 claim that the main response is generated during the preface, \"reducing the delay before the substantive answer,\" is exactly what this missing number would test. The initial-to-main gap is a meaningful perceived-pause metric, but it is not the same as time-to-substantive-answer, and the paper conflates them.\n\nTwo secondary quirks. The non-parametric tests treat response instances as independent even though there are multiple responses per interaction (47–75 interactions per condition). A mixed-effects or cluster-robust analysis would be more defensible, though the large latency differences would likely survive. Condition assignment by time slot with self-selected visitors is weaker than a controlled lab study, but the authors stay cautious.\n\nWhat does the paper do well? The deployment is genuine, the breakdown annotation (fragments, generic questions) is useful, and the trigger-position analysis gives a concrete sense of when early preparation is feasible. The authors also explicitly note that the subjective nulls do not establish equivalence. The intent-readiness threshold of 0.35 is a hand-set deployment choice, not a fitted parameter, so no circularity concern there.\n\nFor someone in HRI or spoken dialogue systems, this is a useful architectural idea worth knowing about, but the central timing trade-off is unsupported without D or M. I'd send it to peer review — it deserves a serious referee — with an expectation of major revision: report absolute main-response onset, control for initial-response duration, and address the repeated-measures issue.","headline":"Missing initial-response duration undermines the claimed latency trade-off; the architecture is plausible but the central comparison is unverified.","tokens_in":8812,"tokens_out":4940,"would_cite":false,"duration_ms":40151,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A two-stage dialogue framework lets a robot speak a context-aware preface early and deliver its main answer sooner.","keywords":["turn-taking","response latency","dialogue robot","incremental generation","voice activity projection","intent readiness detection","conversational fillers","human-robot interaction"],"falsifier":"Record the absolute time from user utterance end to the onset of the main response in both conditions. If the contextual-preface condition does not show a significantly earlier main-response onset than the fixed-filler condition (or does not approach the no-filler main-onset time), the shorter measured gap is an artifact of the metric rather than evidence of parallel generation. The same check could be done by re-analyzing the existing logs.","tokens_in":7898,"feed_emoji":"🤖","tokens_out":3064,"duration_ms":25473,"temperature":0.7,"pith_summary":"The paper argues that a dialogue robot can reduce the time users wait for a substantive answer by preparing a short, context-dependent preface as soon as the user's intent becomes predictable, and speaking it at a moment chosen by a turn-taking model. In a field experiment with a route-guidance robot in a shopping mall, prefaces cut the initial response latency roughly in half compared with no filler, and shortened the pause before the main answer from 0.74 s to 0.43 s compared with fixed fillers, at the cost of a slightly later first word (1.15 s vs 0.94 s). A sympathetic reader would care because it shows a concrete timing trade-off that can be engineered without waiting for fully end-to-end dialogue systems.","feed_headline":"Robot prefaces shorten the pause before the main answer","feed_subtitle":"Field test in a mall: a short prepared preface beats fixed fillers and no filler on time to the substantive response.","key_machinery":"The key mechanism is a two-stage incremental response architecture. An 'intent readiness detector' — a binary classifier trained on pseudo-labeled and human-annotated corpora — decides, from partial speech recognition, whether the user's intent is predictable enough to prepare a preface. A Voice Activity Projection (VAP) model, which predicts when the user will yield the turn, decides when to speak that preface. The preface (under 10 Japanese characters, non-committal) occupies the post-turn silence while the main LLM-based response is generated.","core_discovery":"The central claim is that decoupling prefatory-response preparation from speech onset — preparing a short content-aware preface once the user's intent is predictable, then delivering it when a voice-activity-projection model says the turn is yielding — yields faster access to the substantive answer than either no filler or a fixed filler. In a deployed shopping-mall robot, the contextual preface produced an initial-to-main gap of 0.43 s versus 0.74 s for a fixed filler, while keeping initial latency (1.15 s) far below the no-filler baseline (2.45 s) and only slightly above the fixed filler (0.94 s). The authors interpret the shorter gap as evidence of temporal overlap between main-response g","pith_inferences":["Because the initial-to-main gap is measured from the end of the initial response, and contextual prefaces are longer than fixed fillers, part of the gap reduction may be mechanical: a later end point makes any fixed main-response onset time produce a smaller gap. The paper does not report absolute main-response onset times conditioned on the initial response, so the overlap interpretation is not f","One testable extension is to measure the absolute time from user utterance end to main-response onset, rather than the relative gap, which would separate true generation overlap from mere re-anchoring of the metric.","If the intent-readiness detector is made to trigger earlier, initial latency could approach the fixed-filler level while keeping the shorter gap; the breakdown analysis (20/251 prefaces) suggests the limit is the generation quality of short, grounded prefaces.","The framework is language- and domain-specific as presented; the same two-stage logic is likely to transfer to full-duplex systems, where the preface could be generated even before the user stops speaking."],"forward_implications":["If the framework holds in broader settings, LLM-based robots can deliver substantive answers about 0.9 s sooner than with no filler, without waiting for final recognition.","The separated preparation and speech-onset control makes the generation latency of the main response partially hidden behind speech, so the user perceives a shorter gap.","The intent-readiness detector firing before utterance end in 58.2% of cases means early preparation is often feasible in real interaction, not only in controlled settings.","The trade-off between initial latency and gap suggests designers can tune the trigger threshold to shift perceived responsiveness between first word and full answer.","Independent generation of preface and main response creates a risk of semantic/prosodic discontinuity, which the authors identify as the main bottleneck for quality."],"fun_headline_variants":["Context-aware prefaces cut gap to robot's main answer","Prefaces tailored to user intent shorten wait for robot's real answer","Context-aware preface gives faster path to robot's main response","Real-world robot: context-aware prefaces trim gap to main answer"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The claim that contextual prefaces let the main response be generated during preface speech rests on the measured 'initial-to-main gap,' which is taken from the end of the initial response; since contextual prefaces are longer than fixed fillers, a later end point alone would shrink this gap even if the main-response generation started at the same absolute time.","fun_headline_variants_meta":{"raw":{"variants":["Context-aware prefaces cut gap to robot's main answer","Prefaces tailored to user intent shorten wait for robot's real answer","Context-aware preface gives faster path to robot's main response","Real-world robot: context-aware prefaces trim gap to main answer"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000988,"raw_usage":{"total_tokens":4003,"prompt_tokens":700,"completion_tokens":3303,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":444,"completion_tokens_details":{"reasoning_tokens":3232}},"tokens_in":444,"tokens_out":3303,"duration_ms":23125,"temperature":1.0,"reasoning_tokens":3232,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T03:17:21.298012+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Record the absolute time from user utterance end to the onset of the main response in both conditions. If the contextual-preface condition does not show a significantly earlier main-response onset than the fixed-filler condition (or does not approach the no-filler main-onset time), the shorter measured gap is an artifact of the metric rather than evidence of parallel generation. The same check could be done by re-analyzing the existing logs.","supporting_citations":[],"review_version":1}