{"id":"c212eb38-aa53-4be0-b92f-8b6772d968be","arxiv_id":"2606.01702","paper_version":1,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":5.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":0,"one_line_summary":"KDH-CAD reaches 92.6% accuracy on mechanical part classification with only 250 training samples by integrating foundation models with domain knowledge and minimal data calibration without fine-tuning.","lead":"KDH-CAD combines pretrained foundation models, structured domain knowledge from textbooks, and a small amount of labeled CAD data to perform tasks like mechanical part classification. This hybrid method targets data scarcity in CAD learning by completing concepts in foundation models and calibrating them with minimal real data.","discovery_kind":"new_method","skeptic_critique":{"model":"grok-4.3","headline":"No significant objection identified","rationale":"The reader's UNVERDICTED verdict stems directly from the absence of the full text, which prevents any technical evaluation of the weakest_assumption. Because the abstract alone supplies no concrete equation, algorithm, or experimental protocol that can be shown to be flawed, no load-bearing concern can be raised. The verdict and confidence level therefore remain appropriate.","tokens_in":1806,"tokens_out":263,"duration_ms":14006,"concrete_test":"Obtain the full manuscript and inspect the methods section for the precise mechanism of concept elicitation, the calibration procedure in latent space, and the exact baselines/SOTA comparisons used for the 250-sample and 1,000-sample regimes; confirm whether accuracy is reported on a held-out test set with standard error.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The abstract describes a hybrid framework that elicits CAD concepts from domain knowledge to supplement foundation models and uses minimal labeled data only for latent-space calibration without fine-tuning. No internal inconsistency, hidden assumption, or unsupported logical step is detectable from the given description of the central claim. The performance numbers are presented as empirical outcomes but cannot be stress-tested for confounds, baselines, or implementation details without the full methods and results.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.3","summary":"The manuscript introduces KDH-CAD, a knowledge-data hybrid framework for CAD learning under data scarcity. It integrates pretrained foundation models with structured domain knowledge from textbooks/tutorials to elicit and complete CAD-relevant concepts, then uses a very small amount of labeled CAD data solely for latent-space calibration (without fine-tuning the foundation model). On real-world mechanical part classification, it reports 92.6% accuracy with 250 training samples and 95.8% with 1,000 samples, matching or exceeding state-of-the-art methods that require substantially more data.","tokens_in":1856,"tokens_out":348,"duration_ms":17581,"significance":"If the empirical results are robust, the work offers a practical route to data-efficient CAD learning by treating the problem as knowledge completion and calibration rather than dataset scaling. This could reduce dependence on large authentic or synthetic CAD corpora and broaden applicability in domains where data collection is costly.","major_comments":[{"comment":"Abstract: the central empirical claim (92.6% accuracy at 250 samples, 95.8% at 1,000 samples, matching SOTA with an order of magnitude less data) is presented without any description of the dataset, baselines, implementation of the knowledge-elicitation or calibration steps, or error analysis. These details are load-bearing for evaluating whether the reported numbers support the hybrid-framework thesis.","section":"Abstract"}],"minor_comments":[],"recommendation":"major_revision","confidential_remarks":"The provided manuscript text consists only of the abstract; the absence of methods, results, or figures prevents a full technical assessment. If the full paper exists on arXiv, it should be supplied for review."},"author_rebuttal":{"model":"grok-4.3","summary":"We thank the referee for the constructive feedback and recommendation for major revision. We agree that the abstract would benefit from additional context to better support the central claims. We address the single major comment below.","responses":[{"response":"We acknowledge the validity of this observation. The current abstract prioritizes the high-level thesis and results but omits key experimental context. In the revised manuscript we will expand the abstract (while preserving length constraints) to briefly specify: the real-world mechanical part classification dataset; the SOTA baselines requiring substantially more data; the knowledge-elicitation procedure that extracts and completes CAD concepts from textbooks/tutorials within pretrained foundation models; the latent-space calibration step performed with 250–1,000 labeled samples without any foundation-model fine-tuning; and that detailed error analysis appears in the experiments section. These additions will make the empirical claims more self-contained and directly address the referee’s concern.","revision_made":"yes","referee_comment":"[Abstract] Abstract: the central empirical claim (92.6% accuracy at 250 samples, 95.8% at 1,000 samples, matching SOTA with an order of magnitude less data) is presented without any description of the dataset, baselines, implementation of the knowledge-elicitation or calibration steps, or error analysis. These details are load-bearing for evaluating whether the reported numbers support the hybrid-framework thesis."}],"tokens_in":1327,"tokens_out":303,"duration_ms":16387,"standing_objections":[]},"desk_editor":{"model":"grok-4.3","letter":"The core idea is to stop chasing bigger CAD datasets and instead treat the problem as filling gaps in foundation models with structured textbook knowledge, then tweaking the latent space with a few hundred labeled examples without any fine-tuning. On real mechanical parts they report 92.6% accuracy at 250 samples and 95.8% at 1,000, which they say matches or beats methods that need far more data.\n\nThat framing is the actual novelty. Most prior work either scales synthetic data or fine-tunes large models; here the claim is that domain knowledge can supply the missing CAD concepts and the small labeled set only does calibration. If the experiments are clean, the practical payoff for data-scarce engineering domains is real.\n\nThe abstract is thin on the details that matter. No description of how the domain knowledge is elicited or injected, no baseline tables, no dataset sizes or splits, and no error analysis. Without those it is hard to judge whether the calibration step is doing the work or whether the numbers reflect something simpler like class imbalance or feature leakage. The full paper may contain the methods, but the provided description leaves the central empirical claim uncheckable.\n\nThis is aimed at researchers working on ML for CAD, few-shot learning in geometry, or hybrid knowledge-plus-model systems. A reader looking for concrete low-data recipes would get value if the implementation turns out to be reproducible.\n\nThe problem is worth referee time. The data-scarcity issue in CAD is genuine and the proposed alternative to pure scaling deserves a proper check on methods and controls. I would send it to review rather than desk reject.","headline":"The paper frames CAD learning as knowledge completion using foundation models plus textbook domain knowledge, then calibrates with tiny labeled sets, claiming strong low-data results on mechanical part classification.","tokens_in":2354,"tokens_out":407,"would_cite":false,"duration_ms":21309,"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":"CAD classification reaches over 92 percent accuracy with only 250 labeled samples by completing missing concepts in foundation models using textbook domain knowledge and calibrating them with minimal data.","keywords":["CAD learning","data scarcity","foundation models","domain knowledge","mechanical part classification","knowledge-data hybrid","low-data regimes","data-efficient learning"],"falsifier":"An experiment in which adding the textbook-derived domain knowledge produces no accuracy gain over the foundation model alone on the same 250-sample mechanical-part classification task, or in which the hybrid method falls below 80 percent accuracy while requiring fine-tuning to reach higher numbers.","tokens_in":2688,"feed_emoji":"🔧","tokens_out":822,"duration_ms":21255,"temperature":0.7,"pith_summary":"The paper treats CAD learning as a problem of knowledge completion and calibration rather than one of collecting ever-larger datasets. It shows that domain knowledge from textbooks can supply CAD-relevant concepts that pretrained foundation models express weakly, after which a few hundred labeled examples adjust those concepts in latent space for the specific geometry at hand. This approach avoids any fine-tuning of the foundation model itself. Experiments on mechanical part classification demonstrate that the resulting system matches or exceeds prior state-of-the-art results while using roughly one-tenth the training data. The central motivation is that authentic CAD data remains scarce, so any method that works reliably at small scale changes what is practically feasible in design automation.","feed_headline":"CAD hits 92% accuracy with 250 samples using hybrid knowledge","feed_subtitle":"Textbook concepts fill gaps in foundation models so a few hundred labels suffice to match methods that need far more data","key_machinery":"The KDH-CAD framework, which uses structured domain knowledge to complete CAD concepts in pretrained foundation models and then calibrates those concepts in latent space with a small labeled set without any model fine-tuning.","core_discovery":"KDH-CAD integrates pretrained foundation models, structured domain knowledge from textbooks and tutorials, and a very small amount of labeled CAD data. Domain knowledge elicits and completes CAD-relevant concepts that are weakly expressed or under-represented in the foundation models, while the labeled data calibrates these concepts in the latent space to account for task-specific geometric variability without fine-tuning the foundation model. On real-world mechanical part classification the method reaches 92.6 percent accuracy with 250 training samples and 95.8 percent with 1,000 samples, continuing to improve as more data is added and matching or exceeding state-of-the-art performance that","pith_inferences":["The method could be tested on other CAD tasks such as retrieval or generative modeling to check whether the same calibration step transfers without task-specific redesign.","If textbook knowledge proves sufficient for many sub-domains, practitioners may shift effort from dataset curation toward systematic extraction and structuring of existing engineering documentation.","Because calibration occurs only in latent space, the foundation model remains available for zero-shot or few-shot use on related but unseen CAD problems.","Similar hybrid completion-plus-calibration patterns may apply to other engineering domains that already possess extensive textbook literature but face data-collection bottlenecks."],"forward_implications":["CAD systems can reach competitive performance in low-data regimes that were previously considered impractical.","Performance scales upward as additional labeled samples become available rather than plateauing early.","Reliance on large-scale authentic or synthetic CAD datasets can be substantially reduced while preserving or exceeding prior accuracy levels.","The same knowledge-completion step can be reused across multiple CAD tasks without retraining the underlying foundation model each time."],"fun_headline_variants":["KDH-CAD reaches 92.6% CAD accuracy with 250 samples","Knowledge hybrid reaches 92.6% CAD accuracy with 250 samples","KDH-CAD calibrates knowledge for 92.6% accuracy at 250 samples","Domain knowledge enables 92.6% CAD accuracy at 250 samples"],"cache_read_input_tokens":2112,"weakest_assumption_plain":"Domain knowledge from textbooks and tutorials can reliably elicit and complete CAD-relevant concepts that are weakly expressed in pretrained foundation models, and a very small labeled set can then calibrate those concepts for task-specific geometry without needing to fine-tune the model.","fun_headline_variants_meta":{"raw":{"variants":["KDH-CAD reaches 92.6% CAD accuracy with 250 samples","Knowledge hybrid reaches 92.6% CAD accuracy with 250 samples","KDH-CAD calibrates knowledge for 92.6% accuracy at 250 samples","Domain knowledge enables 92.6% CAD accuracy at 250 samples"]},"model":"grok-4.3","cost_usd":0.008525,"raw_usage":{"total_tokens":3885,"prompt_tokens":735,"num_sources_used":0,"completion_tokens":83,"cost_in_usd_ticks":85249500,"prompt_tokens_details":{"text_tokens":735,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":3067,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":735,"tokens_out":83,"duration_ms":23229,"temperature":1.0,"reasoning_tokens":3067,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-06-28T12:11:35.075088+00:00","model_set":{"reader":"grok-4.3"},"falsifier":"An experiment in which adding the textbook-derived domain knowledge produces no accuracy gain over the foundation model alone on the same 250-sample mechanical-part classification task, or in which the hybrid method falls below 80 percent accuracy while requiring fine-tuning to reach higher numbers.","supporting_citations":[],"review_version":1}