{"query_id": "b7c19f3e4a8d52c0e6f1ab9472d35c8e", "paper_1_id": "85598700", "paper_1_title": "XNect: Real-time Multi-Person 3D Motion Capture with a Single RGB Camera", "paper_2_id": "64848716", "paper_2_title": "XNect: Real-time Multi-person 3D Human Pose Estimation with a Single RGB Camera", "paper_1_publishedDate": "2017-05-01T01:00:00", "paper_2_publishedDate": "2020-04-30T01:00:00", "relation_type": "generalization", "confidence": 0.94, "bridge_topic": "A study real-time monocular RGB 3D human motion capture with temporally coherent skeletal pose output suitable for live character control.", "question": "How can single-person motion capture insights enable robust tracking in crowded real-time scenes?", "hop1_fact": "The earlier paper introduces real-time monocular 3D human pose estimation for a single person using joint 2D–3D prediction and temporal skeleton fitting from one RGB camera.","hop2_fact": "The later paper extends this idea to multi-person scenes by predicting visible joint evidence and reconstructing full 3D poses while handling occlusions and interactions in real time, as described in the multi-stage pipeline (Stage I–III) in the 2020 paper .", "answer": "Single-person real-time skeleton estimation can be extended by predicting partial joint evidence for multiple individuals and reconstructing full poses with temporal fitting under occlusion.", "reasoning_path": "The 2017 paper establishes real-time monocular 3D pose recovery for one subject. The 2020 paper generalizes the same principle to multiple interacting people with occlusion handling and temporal coherence. Combining them yields the insight that single-person real-time capture frameworks can scale to crowded scenes through multi-person reconstruction and temporal modeling."}
{"query_id": "9d4e7a21c5b8f0a36e2c1d94b7a5e8f3", "paper_1_id": "122690614", "paper_1_title": "A Study on Design of a Drone for Transportation in the Agricultural Sector", "paper_2_id": "129391837", "paper_2_title": "A Study On Conceptual Design of a Drone for Transportation in The Agricultural Sector", "paper_1_publishedDate": "2021-10-11T01:00:00", "paper_2_publishedDate": "2022-09-08T01:00:00", "relation_type": "empirical_extension", "confidence": 0.73, "bridge_concept": "10 kg payload agricultural transportation drone conceptual design and simulation validation", "question": "How did the later conceptual drone design strengthen validation of the agricultural transport drone proposed earlier?", "answer": "The later study added more detailed structural and thrust simulations to confirm the feasibility of carrying a 10 kg payload.", "hop1_fact": "It is designed an agricultural transportation drone and verified through simulation that it could lift approximately 10 kg payload.", "hop2_fact": "It is refined the drone concept and performed additional stress, displacement, and thrust analysis to validate the 10 kg payload capability.","reasoning": "The initial 10 kg payload drone design was later strengthened through additional structural and thrust validation confirming improved feasibility. Directly builds on the earlier design by refining and validating the same drone concept using expanded simulations and analysis."}
{"query_id": "c2a8e5f19d374b60a1e9c4d7f2b58a3d", "paper_1_id": "24740302", "paper_1_title": "Proposal Flow", "paper_2_title": "Proposal Flow: Semantic Correspondences from Object Proposals", "paper_2_id": "42849945", "paper_1_publishedDate": "2016-07-08T01:00:00", "paper_2_publishedDate": "2017-03-21T01:00:00", "relation_type": "empirical_extension", "confidence": 0.98, "bridge_concept": "Proposal-flow semantic correspondence using object proposals and local offset matching, evaluated with region-level metrics on proposal-flow benchmarks.", "question": "What method uses locally regularized object-proposal correspondences for semantic flow?", "answer": "The method is proposal flow with local offset matching, and later is extended the evaluation using the PF-WILLOW and PF-PASCAL benchmark suite.", "hop1_fact": "A proposal-based semantic flow framework introduces local offset matching to estimate reliable correspondences from overlapping object proposals and can transform these region matches into a dense flow field.", "hop2_fact": "An extension study that evaluates the same framework on a broader benchmark suite, adding PF-PASCAL alongside PF-WILLOW and reporting more extensive experiments and validation of the generated ground-truth correspondences.", "reasoning": "Proposal flow first establishes semantic correspondences through local offset matching over object proposals, and the later extension validates that same framework more extensively on PF-WILLOW and the newly added PF-PASCAL benchmark. Explicitly states that a preliminary version appeared earlier and then adds a more detailed presentation, a more challenging benchmark based on PASCAL 2011 keypoints, verification of ground-truth generation, and expanded experiments. The core method remains the same, so the relationship is a clear empirical extension rather than a wholly new method."}
{"query_id": "f1b63d9a2c8e47a05d2f1c9b6e3a8d74", "paper_1_id": "156069243", "paper_1_title": "Accelerating Somewhat Homomorphic Evaluation using FPGAs", "paper_2_id": "19531678", "paper_2_title": "Accelerating LTV Based Homomorphic Encryption in Reconfigurable Hardware", "paper_1_publishedDate": "2015-03-28T01:00:00", "paper_2_publishedDate": "2015-08-24T01:00:00", "relation_type": "same_topic_extension", "confidence": 0.83, "bridge_concept": "FPGA-based acceleration of LTV homomorphic encryption using CRT decomposition and NTT-based polynomial multiplication.", "question": "How can large-degree polynomial operations in LTV homomorphic encryption be accelerated to achieve sub-second AES evaluation?", "answer": "By offloading CRT-decomposed polynomial operations to an FPGA that implements an NTT-based multiplier, enabling fast relinearization and sub-second AES evaluation.", "hop1_fact": "Large-degree polynomial multiplication and relinearization in LTV homomorphic encryption can be accelerated by decomposing coefficients with the Chinese Remainder Theorem and performing computations using an NTT-based multiplier on FPGA hardware.", "hop2_fact": "Using this FPGA-based approach, homomorphic evaluation of AES can be reduced to roughly 0.44 seconds per block due to faster polynomial operations and efficient hardware parallelism.", "reasoning": "Both documents present essentially the same technical approach and results, so the relationship does not reflect a meaningful extension. The hops are constructed from the shared technical content without referencing the documents directly. Applying CRT decomposition and NTT-based polynomial multiplication on FPGA hardware enables efficient large-degree operations, which in turn allows AES evaluation in LTV homomorphic encryption to reach roughly 0.44 seconds per block."}
{"query_id": "4e9b2c7a1d5f83e6b0a4c9d2f7e1a358", "paper_1_id": "78417829", "paper_1_title": "On the duration and cost variability of construction activities: an empirical study", "paper_2_title": "Duration and cost variability of construction activities: an empirical study", "paper_2_id": "8257528", "paper_1_publishedDate": "2019-01-01T01:00:00", "paper_2_publishedDate": "2020-01-01T01:00:00", "relation_type": "extension", "confidence": 0.99, "bridge_concept": "Empirical measurement of activity-level duration and cost variability using log ratios and statistical moments to explain project-level delays.", "question": "Why do construction projects experience delays even if activities do not finish late on average?", "answer": "Because high variability in activity durations propagates through project networks, causing delays even when average activity performance is on time.", "hop1_fact": "Construction activity durations have averages close to planned values but exhibit high variability, with standard deviations often exceeding 0.15 in log scale, indicating large dispersion. ", "hop2_fact": "This variability propagates through project networks via mechanisms such as merge event bias, leading to overall project delays and cost overruns despite activities not being late on average. ", "reasoning": "High dispersion in activity durations means that when multiple activities interact in project networks, especially through parallel paths, delays accumulate due to effects like merge event bias. This explains why overall project overruns emerge even though individual activities are not systematically late. Both documents present identical empirical evidence and conclusions, with the later version refining presentation rather than extending the analysis."}