{"noiseShare":0.4974,"topics":[{"claims":[116,189,449,554,533,819,1521,1657,1951],"deltaPp":-1.3988,"id":0,"keywords":["graph","node","neural","networks","classification"],"label":"Graph Neural Networks","labelFull":"Graph Neural Networks: Efficient, Scalable, Robust Learning","papers":2894,"papersByYear":[44,66,157,191,202,294,539,632,769],"rel":-0.2648,"reviewed":true,"share":[5.2821,4.6512,7.0944,7.3631,7.7188,7.7552,7.2848,5.4188,3.8833],"size":7789,"slope":-0.00042},{"claims":[18,67,126,156,224,398,833,1671,2527],"deltaPp":3.3746,"id":1,"keywords":["protein","molecular","molecules","prediction","structures"],"label":"Molecular & Protein Design","labelFull":"Deep Learning for Molecular and Protein Design","papers":2161,"papersByYear":[10,26,58,68,83,149,296,565,906],"rel":2.811,"reviewed":true,"share":[1.2005,1.8323,2.6209,2.6214,3.1716,3.9304,4.0005,4.8444,4.5751],"size":6020,"slope":0.004434},{"claims":[174,323,470,510,477,539,892,1029,1560],"deltaPp":-4.5128,"id":2,"keywords":["adversarial","attacks","robustness","backdoor","examples"],"label":"Adversarial Robustness","labelFull":"Adversarial Robustness: Attacks, Defenses, and Training","papers":2832,"papersByYear":[72,154,200,226,202,237,405,518,818],"rel":-0.5221,"reviewed":true,"share":[8.6435,10.8527,9.0375,8.7124,7.7188,6.2516,5.4737,4.4414,4.1307],"size":5974,"slope":-0.007812},{"claims":[23,45,58,75,80,126,609,1409,1970],"deltaPp":5.1096,"id":3,"keywords":["fine-tuning","evaluation","framework"],"label":"LLM Fine-Tuning and Evaluation","labelFull":"Frameworks for Fine-Tuning and Evaluating Large Language Models","papers":3039,"papersByYear":[17,33,37,49,55,84,402,946,1416],"rel":2.5037,"reviewed":true,"share":[2.0408,2.3256,1.6719,1.889,2.1016,2.2158,5.4332,8.1111,7.1504],"size":4395,"slope":0.007607},{"claims":[72,160,230,268,289,403,598,956,1311],"deltaPp":-0.6483,"id":4,"keywords":["convergence","gradient","optimization","descent","stochastic"],"label":"Gradient-Based Optimization Convergence","labelFull":"Convergence Analysis and Efficiency of Gradient-Based Optimization","papers":2467,"papersByYear":[38,88,135,158,175,229,343,526,775],"rel":-0.1421,"reviewed":true,"share":[4.5618,6.2016,6.1003,6.091,6.687,6.0406,4.6358,4.51,3.9135],"size":4287,"slope":-0.001775},{"claims":[11,43,77,107,111,231,585,1108,1648],"deltaPp":2.7019,"id":5,"keywords":["3d","point","pose","depth","rendering"],"label":"Neural 3D Scene Reconstruction","labelFull":"Neural 3D Scene Reconstruction and Rendering","papers":1757,"papersByYear":[9,22,39,49,48,104,267,470,749],"rel":2.5008,"reviewed":true,"share":[1.0804,1.5504,1.7623,1.889,1.8342,2.7433,3.6086,4.0298,3.7823],"size":3921,"slope":0.003799},{"claims":[88,129,231,267,285,341,549,548,1034],"deltaPp":-2.7761,"id":6,"keywords":["reinforcement","learning","offline","actor-critic","policy"],"label":"Offline Reinforcement Learning","labelFull":"Offline Reinforcement Learning: Algorithms and Benchmarks","papers":2304,"papersByYear":[54,87,140,170,185,216,352,366,734],"rel":-0.4282,"reviewed":true,"share":[6.4826,6.1311,6.3263,6.5536,7.0692,5.6977,4.7574,3.1381,3.7065],"size":3472,"slope":-0.004013},{"claims":[14,13,34,22,59,140,426,889,1864],"deltaPp":4.6965,"id":7,"keywords":["vision-language","visual","reasoning","image","tokens"],"label":"Vision-Language Model Benchmarking","labelFull":"Robust Benchmarking and Evaluation of Vision-Language Models","papers":2108,"papersByYear":[9,11,23,19,36,86,258,522,1144],"rel":4.347,"reviewed":true,"share":[1.0804,0.7752,1.0393,0.7325,1.3756,2.2685,3.487,4.4757,5.7769],"size":3461,"slope":0.006053},{"claims":[23,26,53,68,152,205,507,852,1442],"deltaPp":1.4541,"id":8,"keywords":["series","forecasting","time","multivariate","temporal"],"label":"Time Series Forecasting","labelFull":"Deep Learning for Time Series Forecasting","papers":1382,"papersByYear":[13,13,31,33,63,87,194,351,597],"rel":0.9318,"reviewed":true,"share":[1.5606,0.9161,1.4008,1.2722,2.4073,2.2949,2.622,3.0095,3.0147],"size":3328,"slope":0.002594},{"claims":[31,30,86,141,138,231,542,717,1104],"deltaPp":1.0979,"id":9,"keywords":["causal","counterfactual","treatment","effect","discovery"],"label":"Causal Machine Learning","labelFull":"Causal Inference, Discovery, and Effect Estimation","papers":1474,"papersByYear":[15,17,41,69,67,118,242,331,574],"rel":0.6097,"reviewed":true,"share":[1.8007,1.198,1.8527,2.66,2.5602,3.1126,3.2707,2.838,2.8986],"size":3020,"slope":0.0021},{"claims":[44,90,110,116,132,165,403,762,1139],"deltaPp":0.204,"id":10,"keywords":["speech","audio","music","acoustic","recognition"],"label":"Audio Foundation Models","labelFull":"Audio Foundation Models for Generation and Representation","papers":989,"papersByYear":[15,32,38,43,43,58,128,235,397],"rel":0.1133,"reviewed":true,"share":[1.8007,2.2551,1.7171,1.6577,1.6431,1.5299,1.73,2.0149,2.0047],"size":2961,"slope":-1e-06},{"claims":[24,41,94,101,102,193,350,669,1258],"deltaPp":2.5826,"id":11,"keywords":["state-of-the-art","benchmark","outperforms","baselines","real-world"],"label":"Performance Benchmarking","labelFull":"Empirical Benchmark Evaluation and SOTA Performance Claims","papers":2490,"papersByYear":[24,38,87,91,97,175,320,576,1082],"rel":0.8964,"reviewed":true,"share":[2.8812,2.6779,3.9313,3.5081,3.7065,4.6162,4.3249,4.9387,5.4638],"size":2832,"slope":0.003168},{"claims":[1,3,2,8,49,115,468,862,1291],"deltaPp":3.6572,"id":12,"keywords":["diffusion","denoising","sampling","process","guidance"],"label":"Efficient Diffusion Modeling","labelFull":"Accelerating Training and Sampling of Diffusion Models","papers":1586,"papersByYear":[1,2,2,5,22,66,263,477,748],"rel":30.4767,"reviewed":true,"share":[0.12,0.1409,0.0904,0.1928,0.8407,1.741,3.5545,4.0899,3.7772],"size":2799,"slope":0.005825},{"claims":[45,99,155,263,288,349,515,472,574],"deltaPp":-1.5159,"id":13,"keywords":["label","loss","classification","semi-supervised","noise"],"label":"Learning with Noisy Labels","labelFull":"Robust Semi-supervised Learning with Noisy Labels","papers":1854,"papersByYear":[30,63,110,177,173,222,332,334,413],"rel":-0.4209,"reviewed":true,"share":[3.6014,4.4397,4.9706,6.8234,6.6106,5.856,4.4871,2.8638,2.0855],"size":2760,"slope":-0.002121},{"claims":[2,4,17,18,22,44,167,406,2017],"deltaPp":5.8145,"id":14,"keywords":["reasoning","mathematical","chain-of-thought","math","benchmarks"],"label":"Chain-of-Thought Reasoning","labelFull":"Chain-of-Thought Reasoning for Math and Complex Benchmarks","papers":1635,"papersByYear":[2,4,12,10,13,30,99,266,1199],"rel":24.217,"reviewed":true,"share":[0.2401,0.2819,0.5423,0.3855,0.4968,0.7913,1.338,2.2807,6.0546],"size":2697,"slope":0.005209},{"claims":[18,42,36,44,52,74,200,497,1667],"deltaPp":3.1182,"id":15,"keywords":["agents","ai","llm","systems","benchmark"],"label":"LLM Agents","labelFull":"LLM Agents: Benchmarks, Evaluation, and System Design","papers":1428,"papersByYear":[11,20,21,36,38,44,115,264,879],"rel":2.3614,"reviewed":true,"share":[1.3205,1.4094,0.9489,1.3878,1.452,1.1606,1.5543,2.2636,4.4387],"size":2630,"slope":0.00267},{"claims":[47,89,137,128,155,185,318,454,875],"deltaPp":-0.9161,"id":16,"keywords":["manipulation","robot","imitation","navigation","locomotion"],"label":"Imitation Learning for Robotics","labelFull":"Imitation Learning and Offline RL for Robotic Control","papers":1425,"papersByYear":[29,59,91,93,95,107,186,257,508],"rel":-0.2631,"reviewed":true,"share":[3.4814,4.1579,4.1121,3.5852,3.6301,2.8225,2.5139,2.2035,2.5653],"size":2388,"slope":-0.002248},{"claims":[136,188,222,214,167,238,326,354,422],"deltaPp":-8.3784,"id":17,"keywords":["relu","networks","activation","neural","neurons"],"label":"Overparameterized ReLU Networks","labelFull":"Training Efficiency of Overparameterized ReLU Networks","papers":1512,"papersByYear":[83,121,135,136,114,144,224,241,314],"rel":-0.8409,"reviewed":true,"share":[9.964,8.5271,6.1003,5.2429,4.3561,3.7985,3.0274,2.0664,1.5856],"size":2267,"slope":-0.010081},{"claims":[0,14,24,89,187,314,513,403,535],"deltaPp":1.2725,"id":18,"keywords":["federated","client","learning","personalized","communication"],"label":"Federated Learning","labelFull":"Federated Learning: Efficient and Personalized Methods","papers":912,"papersByYear":[0,7,12,40,73,134,215,179,252],"rel":null,"reviewed":true,"share":[0.0,0.4933,0.5423,1.542,2.7895,3.5347,2.9058,1.5348,1.2725],"size":2079,"slope":0.002489},{"claims":[27,37,57,72,97,108,279,487,890],"deltaPp":0.897,"id":19,"keywords":["attention","heads","transformer","mechanism","linear"],"label":"Efficient Attention","labelFull":"Efficient Attention Mechanisms for Transformers","papers":1267,"papersByYear":[16,22,40,41,48,70,179,293,558],"rel":0.467,"reviewed":true,"share":[1.9208,1.5504,1.8075,1.5806,1.8342,1.8465,2.4192,2.5122,2.8178],"size":2054,"slope":0.001327},{"claims":[3,34,65,106,147,205,292,478,659],"deltaPp":1.4869,"id":20,"keywords":["pde","equations","differential","physics-informed","neural"],"label":"Neural PDE Solvers","labelFull":"Neural PDE Solvers: Physics-Informed Networks and Operators","papers":1006,"papersByYear":[2,19,31,50,68,94,160,240,342],"rel":6.1928,"reviewed":true,"share":[0.2401,1.339,1.4008,1.9275,2.5984,2.4796,2.1625,2.0578,1.727],"size":1989,"slope":0.001697},{"claims":[3,8,23,18,37,55,126,454,1254],"deltaPp":3.9978,"id":21,"keywords":["token","attention","decoding","prediction","reduces"],"label":"Efficient Tokenization","labelFull":"Efficient Tokenization and Pruning for LLMs","papers":1381,"papersByYear":[3,7,16,14,28,40,93,317,863],"rel":11.1019,"reviewed":true,"share":[0.3601,0.4933,0.723,0.5397,1.0699,1.0551,1.2569,2.718,4.3579],"size":1978,"slope":0.004041},{"claims":[25,26,61,100,76,152,280,393,655],"deltaPp":0.1081,"id":22,"keywords":["distillation","student","teacher","knowledge","loss"],"label":"Knowledge Distillation","labelFull":"Knowledge Distillation for Model and Dataset Compression","papers":968,"papersByYear":[15,13,27,41,44,80,153,217,378],"rel":0.06,"reviewed":true,"share":[1.8007,0.9161,1.2201,1.5806,1.6813,2.1103,2.0678,1.8606,1.9088],"size":1768,"slope":0.000915},{"claims":[9,37,54,71,87,133,235,417,655],"deltaPp":1.1128,"id":23,"keywords":["continual","unlearning","forgetting","learning","catastrophic"],"label":"Continual Learning","labelFull":"Continual Learning: Mitigating Catastrophic Forgetting","papers":943,"papersByYear":[6,23,28,41,44,72,128,238,363],"rel":1.5449,"reviewed":true,"share":[0.7203,1.6209,1.2653,1.5806,1.6813,1.8992,1.73,2.0406,1.8331],"size":1698,"slope":0.00116},{"claims":[33,72,125,119,127,169,217,268,463],"deltaPp":-0.8989,"id":24,"keywords":["policy","gradient","optimization","value","algorithm"],"label":"Off-Policy Policy Optimization","labelFull":"Off-Policy Policy Optimization in Reinforcement Learning","papers":1095,"papersByYear":[22,46,81,88,77,111,149,176,345],"rel":-0.3404,"reviewed":true,"share":[2.6411,3.2417,3.6602,3.3924,2.9423,2.928,2.0138,1.509,1.7422],"size":1593,"slope":-0.002092},{"claims":[32,30,38,35,16,44,110,421,860],"deltaPp":0.318,"id":25,"keywords":["question","retrieval-augmented generation","answering","pairs","knowledge"],"label":"Retrieval-Augmented Generation","labelFull":"Retrieval-Augmented Generation for Question Answering","papers":1069,"papersByYear":[22,15,25,21,9,29,83,279,586],"rel":0.1204,"reviewed":true,"share":[2.6411,1.0571,1.1297,0.8096,0.3439,0.765,1.1218,2.3922,2.9591],"size":1586,"slope":0.000869},{"claims":[34,56,88,131,119,138,228,254,344],"deltaPp":-1.538,"id":26,"keywords":["posterior","bayesian","estimator","gaussian","distributions"],"label":"Bayesian Posterior Estimation","labelFull":"Bayesian Posterior Estimation and Approximate Inference","papers":993,"papersByYear":[24,40,59,88,88,97,145,186,266],"rel":-0.5338,"reviewed":true,"share":[2.8812,2.8189,2.6661,3.3924,3.3626,2.5587,1.9597,1.5948,1.3432],"size":1392,"slope":-0.002012},{"claims":[21,25,72,54,74,102,275,301,374],"deltaPp":-0.2752,"id":27,"keywords":["detection","anomaly","out-of-distribution","ood","outlier"],"label":"Anomaly Detection","labelFull":"Anomaly Detection and Out-of-Distribution Benchmarking","papers":632,"papersByYear":[11,12,35,31,34,47,114,141,207],"rel":-0.2084,"reviewed":true,"share":[1.3205,0.8457,1.5816,1.1951,1.2992,1.2398,1.5407,1.209,1.0453],"size":1298,"slope":-8e-06},{"claims":[10,14,18,42,30,40,187,323,619],"deltaPp":1.4993,"id":28,"keywords":["code","instruction","tool","program","test"],"label":"Code Generation Benchmarks","labelFull":"Benchmarks and Datasets for Evaluating Code Generation Models","papers":757,"papersByYear":[4,12,12,20,14,19,98,186,392],"rel":3.1222,"reviewed":true,"share":[0.4802,0.8457,0.5423,0.771,0.535,0.5012,1.3245,1.5948,1.9795],"size":1283,"slope":0.00159},{"claims":[0,0,0,0,6,37,270,427,538],"deltaPp":1.9997,"id":29,"keywords":["diffusion","text-to-image","video","generation","quality"],"label":"Controllable Diffusion","labelFull":"Controllable Diffusion: Alignment, Editing, and Acceleration","papers":908,"papersByYear":[0,0,0,0,5,25,178,304,396],"rel":null,"reviewed":true,"share":[0.0,0.0,0.0,0.0,0.1911,0.6595,2.4057,2.6065,1.9997],"size":1278,"slope":0.003548},{"claims":[26,54,84,62,38,60,125,310,512],"deltaPp":0.3222,"id":30,"keywords":["quantization","quantized","post-training","activation","weight"],"label":"Neural Network Quantization","labelFull":"Post-Training and QAT for Low-Bit Neural Networks","papers":603,"papersByYear":[8,21,35,25,21,31,61,147,254],"rel":0.3355,"reviewed":true,"share":[0.9604,1.4799,1.5816,0.9638,0.8024,0.8177,0.8244,1.2604,1.2826],"size":1271,"slope":-0.000172},{"claims":[9,28,46,32,45,91,183,290,543],"deltaPp":1.147,"id":31,"keywords":["super-resolution","image","restoration","denoising","reconstruction"],"label":"Image Restoration","labelFull":"Image Restoration: Denoising, Super-Resolution, Enhancement","papers":791,"papersByYear":[5,19,25,23,29,54,115,175,346],"rel":1.911,"reviewed":true,"share":[0.6002,1.339,1.1297,0.8867,1.1081,1.4244,1.5543,1.5005,1.7472],"size":1267,"slope":0.001077},{"claims":[24,50,80,107,96,122,238,242,299],"deltaPp":-0.4183,"id":32,"keywords":["few-shot","zero-shot","classification","learning","generalization"],"label":"Few-Shot Learning","labelFull":"Few-Shot Learning for Classification and Zero-Shot Transfer","papers":864,"papersByYear":[14,29,43,54,53,85,152,184,250],"rel":-0.2489,"reviewed":true,"share":[1.6807,2.0437,1.9431,2.0817,2.0252,2.2422,2.0543,1.5776,1.2624],"size":1258,"slope":-0.000448},{"claims":[12,11,43,64,109,155,189,298,359],"deltaPp":-0.0464,"id":33,"keywords":["privacy","private","differential","utility","guarantees"],"label":"Differentially Private Machine Learning","labelFull":"Differential Privacy in ML: Algorithms and Trade-offs","papers":593,"papersByYear":[8,6,18,30,47,69,101,133,181],"rel":-0.0483,"reviewed":true,"share":[0.9604,0.4228,0.8134,1.1565,1.7959,1.8201,1.365,1.1404,0.914],"size":1240,"slope":0.000622},{"claims":[0,16,8,21,27,53,221,295,567],"deltaPp":1.8937,"id":34,"keywords":["multimodal","modalities","fusion","information","representations"],"label":"Missing-Modality Multimodal Learning","labelFull":"Robust Multimodal Learning with Missing Modalities","papers":746,"papersByYear":[0,9,8,13,13,24,120,184,375],"rel":null,"reviewed":true,"share":[0.0,0.6342,0.3615,0.5012,0.4968,0.6331,1.6218,1.5776,1.8937],"size":1208,"slope":0.002176},{"claims":[11,21,55,51,59,130,210,235,287],"deltaPp":0.1343,"id":35,"keywords":["segmentation","semantic","instance","object","detection"],"label":"Image Segmentation","labelFull":"Image Segmentation Models and Benchmark Performance","papers":642,"papersByYear":[7,11,29,32,35,75,120,140,193],"rel":0.1598,"reviewed":true,"share":[0.8403,0.7752,1.3104,1.2336,1.3374,1.9784,1.6218,1.2004,0.9746],"size":1059,"slope":0.00053},{"claims":[7,6,21,32,68,106,202,292,321],"deltaPp":0.3328,"id":36,"keywords":["regret","bandit","bound","contextual","multi-armed"],"label":"Bandit Regret Bounds","labelFull":"Regret Bounds for Multi-Armed and Contextual Bandits","papers":492,"papersByYear":[4,2,10,15,28,52,91,129,161],"rel":0.693,"reviewed":true,"share":[0.4802,0.1409,0.4519,0.5783,1.0699,1.3717,1.2299,1.1061,0.813],"size":1055,"slope":0.001096},{"claims":[8,13,12,8,30,40,106,300,530],"deltaPp":0.4653,"id":37,"keywords":["eeg","brain","fmri","signals","decoding"],"label":"EEG Foundation Models","labelFull":"Large-Scale Pretrained EEG Foundation Models for Decoding","papers":445,"papersByYear":[5,6,8,6,16,20,52,121,211],"rel":0.7752,"reviewed":true,"share":[0.6002,0.4228,0.3615,0.2313,0.6114,0.5276,0.7028,1.0375,1.0655],"size":1047,"slope":0.000781},{"claims":[15,13,28,18,21,32,141,269,505],"deltaPp":0.8817,"id":38,"keywords":["retrieval","query","documents","embeddings","information"],"label":"Retrieval-Augmented Generation","labelFull":"Retrieval-Augmented Generation for Large Language Models","papers":689,"papersByYear":[7,10,22,13,17,26,81,172,341],"rel":1.0493,"reviewed":true,"share":[0.8403,0.7047,0.9941,0.5012,0.6496,0.6858,1.0947,1.4747,1.722],"size":1042,"slope":0.001037},{"claims":[15,13,23,28,15,23,117,297,492],"deltaPp":1.0662,"id":39,"keywords":["video","motion","generation","frames","temporal"],"label":"Long Video Generation","labelFull":"Long-Form Video Generation with Temporal Consistency","papers":670,"papersByYear":[5,10,17,20,8,17,75,188,330],"rel":1.7764,"reviewed":true,"share":[0.6002,0.7047,0.7682,0.771,0.3057,0.4484,1.0137,1.6119,1.6664],"size":1023,"slope":0.001192},{"claims":[15,9,39,55,77,130,193,195,246],"deltaPp":-0.586,"id":40,"keywords":["vision","image classification","transformers","visual","recognition"],"label":"Vision Transformers","labelFull":"Vision Transformers for Efficient Image Classification","papers":744,"papersByYear":[13,9,31,46,57,95,146,154,193],"rel":-0.3755,"reviewed":true,"share":[1.5606,0.6342,1.4008,1.7733,2.1781,2.5059,1.9732,1.3204,0.9746],"size":959,"slope":0.000265},{"claims":[2,7,16,33,33,83,157,259,361],"deltaPp":1.0122,"id":41,"keywords":["transformer","layer","trained","architectures","enables"],"label":"Efficient Transformers","labelFull":"Efficient Transformer Architectures and Training Techniques","papers":639,"papersByYear":[2,4,15,22,24,56,96,172,248],"rel":4.2157,"reviewed":true,"share":[0.2401,0.2819,0.6778,0.8481,0.9171,1.4772,1.2975,1.4747,1.2523],"size":951,"slope":0.001583},{"claims":[20,12,46,26,39,68,182,201,340],"deltaPp":-0.1676,"id":42,"keywords":["spiking","neural","networks","neuron","synaptic"],"label":"Spiking Neural Network Training","labelFull":"Efficient Training Methods for Spiking Neural Networks","papers":430,"papersByYear":[8,9,19,15,19,29,77,97,157],"rel":-0.1745,"reviewed":true,"share":[0.9604,0.6342,0.8586,0.5783,0.726,0.765,1.0407,0.8317,0.7928],"size":934,"slope":7.9e-05},{"claims":[7,7,2,6,8,31,161,219,457],"deltaPp":1.1963,"id":43,"keywords":["editing","text-to-image generation","video","prompts","quality"],"label":"Text-to-Image Benchmarks","labelFull":"Benchmarks and Metrics for Text-to-Image Generation","papers":642,"papersByYear":[4,4,2,4,7,19,108,162,332],"rel":2.4913,"reviewed":true,"share":[0.4802,0.2819,0.0904,0.1542,0.2675,0.5012,1.4597,1.389,1.6765],"size":898,"slope":0.001865},{"claims":[6,7,11,17,17,37,68,232,489],"deltaPp":1.2541,"id":44,"keywords":["inference","cache","kv","memory","compression"],"label":"KV-Cache Compression","labelFull":"KV-Cache Compression for Efficient LLM Inference","papers":729,"papersByYear":[6,6,11,17,15,35,64,184,391],"rel":1.7411,"reviewed":true,"share":[0.7203,0.4228,0.4971,0.6554,0.5732,0.9232,0.865,1.5776,1.9744],"size":884,"slope":0.001581},{"claims":[7,14,20,41,41,54,100,212,383],"deltaPp":0.7531,"id":45,"keywords":["low-rank","matrix","matrices","singular","factorization"],"label":"Low-Rank Adaptation","labelFull":"Low-Rank Adaptation for Parameter-Efficient Fine-Tuning","papers":584,"papersByYear":[5,10,16,24,24,33,63,141,268],"rel":1.2547,"reviewed":true,"share":[0.6002,0.7047,0.723,0.9252,0.9171,0.8705,0.8515,1.209,1.3533],"size":872,"slope":0.000788},{"claims":[6,18,33,74,61,50,110,136,364],"deltaPp":0.743,"id":46,"keywords":["uncertainty","calibration","epistemic","quantification","confidence"],"label":"Uncertainty Quantification","labelFull":"Uncertainty Quantification and Calibration in Machine Learning","papers":624,"papersByYear":[5,13,17,49,42,43,84,105,266],"rel":1.2379,"reviewed":true,"share":[0.6002,0.9161,0.7682,1.889,1.6049,1.1343,1.1353,0.9003,1.3432],"size":852,"slope":0.000484},{"claims":[11,20,52,44,62,134,193,132,196],"deltaPp":-0.1485,"id":47,"keywords":["domain","adaptation","generalization","unsupervised","target"],"label":"Unsupervised Domain Adaptation","labelFull":"Unsupervised Domain Adaptation and Generalization","papers":517,"papersByYear":[7,12,26,25,42,71,114,83,137],"rel":-0.1767,"reviewed":true,"share":[0.8403,0.8457,1.1749,0.9638,1.6049,1.8729,1.5407,0.7117,0.6918],"size":844,"slope":0.000107},{"claims":[3,6,7,8,8,9,74,216,495],"deltaPp":1.2154,"id":48,"keywords":["human","evaluation","llm","judges","responses"],"label":"LLM-as-a-Judge Evaluation","labelFull":"LLM-as-a-Judge Evaluation and Human Alignment","papers":537,"papersByYear":[3,5,7,6,5,9,53,137,312],"rel":3.3752,"reviewed":true,"share":[0.3601,0.3524,0.3163,0.2313,0.1911,0.2374,0.7163,1.1747,1.5755],"size":826,"slope":0.001356},{"claims":[4,3,12,11,12,13,116,173,481],"deltaPp":1.3983,"id":49,"keywords":["reinforcement","agents","learning","reward","feedback"],"label":"Reinforcement Learning for LLMs","labelFull":"Reinforcement Learning for Large Language Model Agents","papers":617,"papersByYear":[4,1,7,11,6,9,80,127,372],"rel":2.9119,"reviewed":true,"share":[0.4802,0.0705,0.3163,0.4241,0.2293,0.2374,1.0812,1.0889,1.8785],"size":825,"slope":0.001665},{"claims":[0,0,2,2,5,49,158,223,331],"deltaPp":1.2927,"id":50,"keywords":["prompt","tuning","optimization","multiple","learning"],"label":"Prompt Optimization","labelFull":"Automatic Prompt Optimization for Large Language Models","papers":569,"papersByYear":[0,0,1,1,5,31,114,161,256],"rel":null,"reviewed":true,"share":[0.0,0.0,0.0452,0.0386,0.1911,0.8177,1.5407,1.3804,1.2927],"size":770,"slope":0.00218},{"claims":[38,94,89,80,81,62,96,102,109],"deltaPp":-2.0716,"id":51,"keywords":["variational","autoencoders","latent","inference","posterior"],"label":"Variational Autoencoders","labelFull":"Latent Variable Modeling with Variational Autoencoders","papers":521,"papersByYear":[21,60,64,54,50,51,63,69,89],"rel":-0.8217,"reviewed":true,"share":[2.521,4.2283,2.892,2.0817,1.9106,1.3453,0.8515,0.5916,0.4494],"size":751,"slope":-0.004002},{"claims":[43,42,47,52,56,57,81,135,230],"deltaPp":-2.7772,"id":52,"keywords":["generative","discriminative","modeling","deep","latent"],"label":"Plug-and-Play Generative Models","labelFull":"Plug-and-Play Generative Modeling and Sampling Frameworks","papers":577,"papersByYear":[31,30,33,41,43,39,69,104,187],"rel":-0.7463,"reviewed":true,"share":[3.7215,2.1142,1.4912,1.5806,1.6431,1.0288,0.9326,0.8917,0.9443],"size":743,"slope":-0.002741},{"claims":[11,16,31,34,48,45,101,152,291],"deltaPp":0.3227,"id":53,"keywords":["recommendation","user","item","personalization","collaborative filtering"],"label":"Personalized Recommendation","labelFull":"Personalized Recommendation and Model-Agnostic Enhancements","papers":362,"papersByYear":[4,7,14,17,21,22,47,71,159],"rel":0.672,"reviewed":true,"share":[0.4802,0.4933,0.6326,0.6554,0.8024,0.5803,0.6352,0.6088,0.8029],"size":729,"slope":0.000261},{"claims":[0,8,4,5,6,9,119,196,377],"deltaPp":0.5605,"id":54,"keywords":["watermark","robustness","removal","text","attacks"],"label":"Robust Generative Watermarking","labelFull":"Robust Watermarking for LLMs and Diffusion Models","papers":220,"papersByYear":[0,3,2,2,3,5,33,61,111],"rel":null,"reviewed":true,"share":[0.0,0.2114,0.0904,0.0771,0.1146,0.1319,0.446,0.523,0.5605],"size":724,"slope":0.000657},{"claims":[3,21,8,17,20,32,100,222,291],"deltaPp":0.4719,"id":55,"keywords":["driving","traffic","autonomous","road","vehicle"],"label":"Closed-Loop Autonomous Driving Benchmarks","labelFull":"Closed-Loop Benchmarks for Autonomous Driving","papers":345,"papersByYear":[2,9,8,11,11,19,44,100,141],"rel":1.9654,"reviewed":true,"share":[0.2401,0.6342,0.3615,0.4241,0.4203,0.5012,0.5947,0.8574,0.712],"size":714,"slope":0.000517},{"claims":[33,61,86,86,64,74,105,96,108],"deltaPp":-2.8412,"id":56,"keywords":["cifar-10","cifar-100","imagenet","mnist","cifar10"],"label":"CIFAR-10/100 and ImageNet Classification","labelFull":"State-of-the-Art on CIFAR-10/100 and ImageNet Classification","papers":677,"papersByYear":[28,58,83,83,61,69,100,92,103],"rel":-0.8453,"reviewed":true,"share":[3.3613,4.0874,3.7506,3.1997,2.3309,1.8201,1.3515,0.7888,0.5201],"size":713,"slope":-0.004573},{"claims":[6,23,54,54,34,73,95,120,238],"deltaPp":0.0967,"id":57,"keywords":["knowledge graph","entity","relation","graph completion","embedding"],"label":"Knowledge Graph Completion","labelFull":"Knowledge Graph Completion and Embedding Techniques","papers":379,"papersByYear":[5,13,27,28,16,34,46,72,138],"rel":0.1611,"reviewed":true,"share":[0.6002,0.9161,1.2201,1.0794,0.6114,0.8969,0.6217,0.6173,0.6969],"size":697,"slope":-0.000315},{"claims":[106,161,116,95,72,54,41,29,22],"deltaPp":-5.3265,"id":58,"keywords":["gan","generator","discriminator","wasserstein","samples"],"label":"GAN Training Stabilization","labelFull":"Methods for Stabilizing and Improving GAN Training","papers":335,"papersByYear":[45,68,56,41,37,28,23,22,15],"rel":-0.986,"reviewed":true,"share":[5.4022,4.7921,2.5305,1.5806,1.4138,0.7386,0.3109,0.1886,0.0757],"size":696,"slope":-0.006733},{"claims":[17,19,32,30,55,55,116,156,195],"deltaPp":-0.0245,"id":59,"keywords":["explanations","explainability","faithfulness","metrics","human"],"label":"Post-Hoc Explanation Methods","labelFull":"Post-Hoc Explanation Methods for Black-Box Models","papers":383,"papersByYear":[5,11,18,18,28,33,65,91,114],"rel":-0.0408,"reviewed":true,"share":[0.6002,0.7752,0.8134,0.6939,1.0699,0.8705,0.8785,0.7802,0.5757],"size":675,"slope":3.7e-05},{"claims":[20,34,53,72,60,64,89,97,179],"deltaPp":-0.6085,"id":60,"keywords":["multi-task","transfer","task-specific","learning","knowledge"],"label":"Multi-Task Learning","labelFull":"Multi-Task Learning: Joint Training, Meta-Learning, Transfer","papers":500,"papersByYear":[11,24,39,49,42,48,68,78,141],"rel":-0.4608,"reviewed":true,"share":[1.3205,1.6913,1.7623,1.889,1.6049,1.2662,0.919,0.6688,0.712],"size":668,"slope":-0.001302},{"claims":[5,17,47,35,72,51,109,144,186],"deltaPp":0.1056,"id":61,"keywords":["equivariant","group","symmetries","topological","invariant"],"label":"Group Equivariant Neural Networks","labelFull":"Group Equivariant Neural Networks: Theory and Architectures","papers":375,"papersByYear":[4,7,18,22,36,31,60,81,116],"rel":0.2199,"reviewed":true,"share":[0.4802,0.4933,0.8134,0.8481,1.3756,0.8177,0.8109,0.6945,0.5858],"size":666,"slope":0.000165},{"claims":[0,0,0,4,2,8,43,264,339],"deltaPp":1.0958,"id":62,"keywords":["preference","optimization","direct","alignment","dpo"],"label":"Preference Optimization","labelFull":"Preference Optimization Techniques for LLM Alignment","papers":417,"papersByYear":[0,0,0,3,2,4,33,158,217],"rel":null,"reviewed":true,"share":[0.0,0.0,0.0,0.1157,0.0764,0.1055,0.446,1.3547,1.0958],"size":660,"slope":0.001555},{"claims":[7,9,30,46,42,96,95,145,184],"deltaPp":-0.0106,"id":63,"keywords":["games","equilibrium","nash","equilibria","zero-sum"],"label":"Nash Equilibrium Learning","labelFull":"Algorithms For Learning Nash Equilibria In Multi-Agent Games","papers":317,"papersByYear":[4,7,13,20,20,42,49,69,93],"rel":-0.0221,"reviewed":true,"share":[0.4802,0.4933,0.5874,0.771,0.7642,1.1079,0.6623,0.5916,0.4696],"size":654,"slope":0.000123},{"claims":[1,6,11,20,16,20,67,163,347],"deltaPp":1.0717,"id":64,"keywords":["reward","preference","functions","modeling","human"],"label":"Reward Modeling","labelFull":"Reward Modeling for Reinforcement Learning from Human Feedback","papers":457,"papersByYear":[1,5,11,15,16,14,45,114,236],"rel":8.9308,"reviewed":true,"share":[0.12,0.3524,0.4971,0.5783,0.6114,0.3693,0.6082,0.9775,1.1917],"size":651,"slope":0.001029},{"claims":[9,16,31,26,43,48,94,130,242],"deltaPp":-0.1648,"id":65,"keywords":["face","facial","recognition","talking","identity"],"label":"Identity-Preserving Face Synthesis","labelFull":"Identity-Preserving Face Synthesis, Editing, and Recognition","papers":295,"papersByYear":[6,12,16,17,16,17,46,55,110],"rel":-0.2288,"reviewed":true,"share":[0.7203,0.8457,0.723,0.6554,0.6114,0.4484,0.6217,0.4716,0.5555],"size":639,"slope":-0.000365},{"claims":[18,28,21,23,4,23,75,132,311],"deltaPp":-0.0262,"id":66,"keywords":["dialogue","conversational","multi-turn","spoken","user"],"label":"Long‑Form Dialogue","labelFull":"Benchmarks and Methods for Long‑Form Dialogue","papers":355,"papersByYear":[8,12,12,8,3,10,43,74,185],"rel":-0.0273,"reviewed":true,"share":[0.9604,0.8457,0.5423,0.3084,0.1146,0.2638,0.5812,0.6345,0.9342],"size":635,"slope":-0.000118},{"claims":[0,3,3,4,0,1,45,184,389],"deltaPp":1.0049,"id":67,"keywords":["hallucination","vision-language","reduces","detection","mitigates"],"label":"Multimodal Hallucination Mitigation","labelFull":"Hallucination Detection and Mitigation in Multimodal Models","papers":320,"papersByYear":[0,1,2,2,0,1,22,93,199],"rel":null,"reviewed":true,"share":[0.0,0.0705,0.0904,0.0771,0.0,0.0264,0.2973,0.7974,1.0049],"size":629,"slope":0.001094},{"claims":[20,35,35,27,58,31,67,125,215],"deltaPp":-1.1381,"id":68,"keywords":["compression","image","techniques","learned","neural"],"label":"Neural Network Compression","labelFull":"Neural Network Compression for Efficient Inference","papers":412,"papersByYear":[16,23,21,19,31,21,43,83,155],"rel":-0.5925,"reviewed":true,"share":[1.9208,1.6209,0.9489,0.7325,1.1846,0.5539,0.5812,0.7117,0.7827],"size":613,"slope":-0.001366},{"claims":[0,4,16,42,45,100,157,104,141],"deltaPp":0.3939,"id":69,"keywords":["fairness","sensitive","attributes","group","metrics"],"label":"Fairness-Accuracy Tradeoff","labelFull":"Fairness-Accuracy Tradeoff Optimization and Evaluation","papers":291,"papersByYear":[0,2,7,18,21,41,72,52,78],"rel":null,"reviewed":true,"share":[0.0,0.1409,0.3163,0.6939,0.8024,1.0815,0.9731,0.4459,0.3939],"size":609,"slope":0.000699},{"claims":[4,5,20,39,47,58,105,106,205],"deltaPp":0.247,"id":70,"keywords":["detection","object","detectors","coco","state-of-the-art"],"label":"Object Detection","labelFull":"Object Detection: Benchmarks, OOD, and Emerging Paradigms","papers":388,"papersByYear":[4,4,15,23,28,34,63,73,144],"rel":0.5144,"reviewed":true,"share":[0.4802,0.2819,0.6778,0.8867,1.0699,0.8969,0.8515,0.6259,0.7272],"size":589,"slope":0.000396},{"claims":[79,82,74,44,23,33,57,99,91],"deltaPp":-3.7281,"id":71,"keywords":["word","sentence","embeddings","text","similarity"],"label":"Word and Sentence Embeddings","labelFull":"Learning Word and Sentence Embeddings for Text Similarity","papers":350,"papersByYear":[34,40,38,32,14,24,38,60,70],"rel":-0.9134,"reviewed":true,"share":[4.0816,2.8189,1.7171,1.2336,0.535,0.6331,0.5136,0.5144,0.3535],"size":582,"slope":-0.004139},{"claims":[41,66,70,80,53,46,62,64,88],"deltaPp":-2.3937,"id":72,"keywords":["translation","machine","non-autoregressive","multilingual","english"],"label":"Neural Machine Translation","labelFull":"Neural Machine Translation: Models, Benchmarks, Low-Resource","papers":300,"papersByYear":[22,34,42,42,19,25,35,32,49],"rel":-0.9063,"reviewed":true,"share":[2.6411,2.3961,1.8979,1.6191,0.726,0.6595,0.473,0.2744,0.2474],"size":570,"slope":-0.003292},{"claims":[0,0,2,1,3,7,54,132,356],"deltaPp":1.4796,"id":73,"keywords":["reasoning","capabilities","enhances","benchmarks","framework"],"label":"Reasoning-Augmented LLMs","labelFull":"Techniques for Enhancing Reasoning in Large Language Models","papers":449,"papersByYear":[0,0,2,1,3,4,39,107,293],"rel":null,"reviewed":true,"share":[0.0,0.0,0.0904,0.0386,0.1146,0.1055,0.5271,0.9174,1.4796],"size":555,"slope":0.001602},{"claims":[8,13,22,20,13,47,84,119,219],"deltaPp":0.0893,"id":74,"keywords":["quantum","classical","circuits","states","algorithm"],"label":"Quantum Machine Learning","labelFull":"Quantum Machine Learning: Neural Nets, Kernels, Variational","papers":204,"papersByYear":[3,4,5,7,6,16,31,43,89],"rel":0.248,"reviewed":true,"share":[0.3601,0.2819,0.2259,0.2699,0.2293,0.4221,0.419,0.3687,0.4494],"size":545,"slope":0.000193},{"claims":[9,5,6,11,19,37,86,132,238],"deltaPp":0.2571,"id":75,"keywords":["medical","segmentation","mri","imaging","lesion"],"label":"Medical Image Segmentation","labelFull":"Medical Image Segmentation: Models, Datasets, and Benchmarks","papers":332,"papersByYear":[4,5,6,9,12,18,51,81,146],"rel":0.5354,"reviewed":true,"share":[0.4802,0.3524,0.2711,0.347,0.4585,0.4748,0.6893,0.6945,0.7373],"size":543,"slope":0.000503},{"claims":[16,27,30,49,40,57,65,108,146],"deltaPp":-0.0952,"id":76,"keywords":["manifold","riemannian","kernel","space","intrinsic"],"label":"Manifold-Aware Deep Learning","labelFull":"Manifold-Aware Deep Learning and Riemannian Optimization","papers":338,"papersByYear":[5,15,17,28,26,34,47,66,100],"rel":-0.1586,"reviewed":true,"share":[0.6002,1.0571,0.7682,1.0794,0.9935,0.8969,0.6352,0.5659,0.505],"size":538,"slope":-0.000384},{"claims":[10,20,38,33,22,44,90,124,152],"deltaPp":-0.3696,"id":77,"keywords":["text","generation","nlp","natural","understanding"],"label":"Long-Form Text Generation","labelFull":"Improving and Evaluating Long-Form Text Generation","papers":399,"papersByYear":[8,12,26,25,20,34,62,95,117],"rel":-0.3848,"reviewed":true,"share":[0.9604,0.8457,1.1749,0.9638,0.7642,0.8969,0.838,0.8145,0.5908],"size":533,"slope":-0.000385},{"claims":[6,2,10,3,13,12,56,169,259],"deltaPp":0.4237,"id":78,"keywords":["attacks","jailbreak","privacy","adversarial","success"],"label":"LLM Jailbreak Attacks","labelFull":"Jailbreak Attacks and Defenses for Large Language Models","papers":350,"papersByYear":[4,2,6,3,10,7,40,99,179],"rel":0.8823,"reviewed":true,"share":[0.4802,0.1409,0.2711,0.1157,0.3821,0.1846,0.5406,0.8488,0.9039],"size":530,"slope":0.000738},{"claims":[0,1,2,4,6,10,19,140,338],"deltaPp":1.015,"id":79,"keywords":["experts","router","routing","mixture","moe"],"label":"Mixture-of-Experts Routing","labelFull":"Routing, Scaling, and Efficiency for Mixture-of-Experts","papers":321,"papersByYear":[0,1,2,3,4,8,14,88,201],"rel":null,"reviewed":true,"share":[0.0,0.0705,0.0904,0.1157,0.1528,0.211,0.1892,0.7545,1.015],"size":520,"slope":0.001067},{"claims":[61,82,57,66,27,23,53,56,85],"deltaPp":-3.8089,"id":80,"keywords":["recurrent","networks","neural","lstm","memory"],"label":"Efficient Recurrent Networks","labelFull":"Efficient Recurrent Neural Network Architectures","papers":292,"papersByYear":[34,38,29,33,20,19,32,33,54],"rel":-0.9332,"reviewed":true,"share":[4.0816,2.6779,1.3104,1.2722,0.7642,0.5012,0.4325,0.2829,0.2727],"size":510,"slope":-0.004158},{"claims":[23,23,52,54,59,70,90,53,69],"deltaPp":-1.8226,"id":81,"keywords":["imagenet","top-1","accuracy","resnet-50","pruning"],"label":"Efficient ImageNet Models","labelFull":"Efficient ImageNet Models For High Top‑1 Accuracy","papers":435,"papersByYear":[18,21,43,50,53,58,75,50,67],"rel":-0.8434,"reviewed":true,"share":[2.1609,1.4799,1.9431,1.9275,2.0252,1.5299,1.0137,0.4287,0.3383],"size":493,"slope":-0.002117},{"claims":[3,1,8,5,13,12,52,126,244],"deltaPp":0.6044,"id":82,"keywords":["alignment","llm","outperforms","compared","benchmarks"],"label":"LLM Alignment","labelFull":"Alignment Methods for Large Language Models","papers":357,"papersByYear":[3,1,7,4,9,9,39,94,191],"rel":1.6784,"reviewed":true,"share":[0.3601,0.0705,0.3163,0.1542,0.3439,0.2374,0.5271,0.806,0.9645],"size":464,"slope":0.000855},{"claims":[0,1,5,23,13,23,30,107,253],"deltaPp":0.7676,"id":83,"keywords":["flow","matching","velocity","normalizing","field"],"label":"Flow Matching","labelFull":"Flow Matching for Continuous Normalizing Flow Training","papers":270,"papersByYear":[0,1,5,10,10,12,20,60,152],"rel":null,"reviewed":true,"share":[0.0,0.0705,0.2259,0.3855,0.3821,0.3165,0.2703,0.5144,0.7676],"size":455,"slope":0.000737},{"claims":[4,5,2,10,4,9,46,139,222],"deltaPp":0.5287,"id":84,"keywords":["question","answering","visual","video","pairs"],"label":"Visual Question Answering","labelFull":"Visual Question Answering: Datasets, Models, and Evaluation","papers":331,"papersByYear":[3,4,1,7,4,7,32,97,176],"rel":1.4682,"reviewed":true,"share":[0.3601,0.2819,0.0452,0.2699,0.1528,0.1846,0.4325,0.8317,0.8888],"size":441,"slope":0.000742},{"claims":[0,0,0,0,0,0,18,158,265],"deltaPp":1.0099,"id":85,"keywords":["multimodal","evaluation","open-source","benchmark","reveals"],"label":"Multimodal LLM Benchmarks","labelFull":"Benchmarks for Evaluating Multimodal LLMs","papers":317,"papersByYear":[0,0,0,0,0,0,13,104,200],"rel":null,"reviewed":true,"share":[0.0,0.0,0.0,0.0,0.0,0.0,0.1757,0.8917,1.0099],"size":441,"slope":0.001178},{"claims":[17,3,27,35,18,50,66,95,126],"deltaPp":-0.5918,"id":86,"keywords":["clustering","k-means","kernel","deep","algorithm"],"label":"Scalable Clustering","labelFull":"Scalable and Efficient Clustering Algorithms","papers":243,"papersByYear":[8,3,14,20,10,22,37,56,73],"rel":-0.6162,"reviewed":true,"share":[0.9604,0.2114,0.6326,0.771,0.3821,0.5803,0.5001,0.4802,0.3686],"size":437,"slope":-0.000336},{"claims":[4,3,8,6,5,17,25,117,230],"deltaPp":0.4529,"id":87,"keywords":["decoding","speculative","codes","verification","parallel"],"label":"Speculative Decoding","labelFull":"Speculative Decoding for Efficient LLM Inference","papers":293,"papersByYear":[3,1,7,5,4,14,18,80,161],"rel":1.2577,"reviewed":true,"share":[0.3601,0.0705,0.3163,0.1928,0.1528,0.3693,0.2433,0.6859,0.813],"size":415,"slope":0.000615},{"claims":[8,10,16,22,18,28,69,112,128],"deltaPp":-0.3443,"id":88,"keywords":["pre-training","reduces","pretraining","fine-tuning","compared"],"label":"Parameter-Efficient Fine-Tuning","labelFull":"Parameter-Efficient Fine-Tuning of Large Language Models","papers":384,"papersByYear":[8,10,15,20,17,26,64,102,122],"rel":-0.3585,"reviewed":true,"share":[0.9604,0.7047,0.6778,0.771,0.6496,0.6858,0.865,0.8746,0.6161],"size":411,"slope":-9.6e-05},{"claims":[19,16,24,30,15,26,45,77,139],"deltaPp":-0.8851,"id":89,"keywords":["embedding","space","learned","representations","feature"],"label":"Embedding Learning","labelFull":"Learning and Optimizing Embedding Representations","papers":304,"papersByYear":[12,11,19,20,13,21,32,66,110],"rel":-0.6144,"reviewed":true,"share":[1.4406,0.7752,0.8586,0.771,0.4968,0.5539,0.4325,0.5659,0.5555],"size":391,"slope":-0.000873}],"totals":{"claims":[3913,6699,10359,12315,12624,18240,35999,58053,102290],"papers":[833,1419,2213,2594,2617,3791,7399,11663,19803]},"years":[2018,2019,2020,2021,2022,2023,2024,2025,2026]}