NVIDIA \u7814\u7a76\u4eba\u54e1\u5c07\u5728\u672c\u6b21\u5927\u6703\u4e0a\u767c\u8868\u8d85\u904e 70 \u7bc7\u8ad6\u6587\u3001\u8209\u8fa6\u6f14\u8b1b\u548c\u7814\u8a0e\u6703\uff0c\u5206\u4eab\u6db5\u84cb AI \u63a8\u7406\u3001\u91ab\u5b78\u7814\u7a76\u3001\u81ea\u99d5\u8eca\uff08AV\uff09\u958b\u767c\u7b49\u9818\u57df\u7684\u5275\u65b0\u5c08\u6848\u3002<\/p>\n\n\n\n
\u9019\u4e9b\u8a08\u756b\u6df1\u5316\u4e86 NVIDIA \u5c0d\u958b\u6e90\u7684\u627f\u8afe\uff0c\u540c\u6642\u66f4\u7372\u5f97 AI \u57fa\u6e96\u6e2c\u8a66\u6a5f\u69cb Artificial Analysis\u6700\u65b0\u767c\u5e03\u7684Openness Index<\/a> \u8a8d\u53ef\u3002\u8a72\u6307\u6578\u6839\u64da\u6a21\u578b\u6388\u6b0a\u7684\u5408\u6cd5\u6027\u3001\u8cc7\u6599\u900f\u660e\u5ea6\u548c\u6280\u8853\u7d30\u7bc0\u7684\u53ef\u7528\u6027\uff0c\u5c07 NVIDIA Nemotron<\/a> \u7cfb\u5217\u7528\u65bc\u9802\u5c16 AI \u958b\u767c\u7684\u958b\u653e\u6280\u8853\uff0c\u8a55\u70ba AI \u751f\u614b\u7cfb\u4e2d\u6700\u958b\u653e\u7684\u6280\u8853\u4e4b\u4e00\u3002<\/p>\n\n\n\n <\/p>\n\n\n\n NVIDIA DRIVE Alpamayo-R1\uff08AR1\uff09<\/a>\u662f\u5168\u7403\u9996\u6b3e\u7528\u65bc\u81ea\u52d5\u99d5\u99db\u7814\u7a76\u7684\u958b\u653e\u5f0f\u63a8\u7406\u8996\u89ba\u8a9e\u8a00\u52d5\u4f5c\uff08VLA\uff09\u6a21\u578b<\/a>\uff0c\u8a72\u6a21\u578b\u5c07\u601d\u7dad\u93c8\uff08chain-of-thought\uff09AI\u63a8\u7406<\/a>\u8207\u8def\u5f91\u898f\u5283\u76f8\u7d50\u5408\u3002\u8def\u5f91\u898f\u5283\u662f\u63d0\u5347\u81ea\u99d5\u8eca\u5728\u8907\u96dc\u9053\u8def\u5834\u666f\u4e2d\u7684\u5b89\u5168\u6027\uff0c\u4e26\u5be6\u73fe Level 4\u81ea\u52d5\u99d5\u99db<\/a>\u7684\u95dc\u9375\u8981\u7d20\u3002<\/p>\n\n\n\n \u4ee5\u5f80\u7684\u81ea\u52d5\u99d5\u99db\u6a21\u578b\u5728\u8655\u7406\u884c\u4eba\u5bc6\u96c6\u7684\u5341\u5b57\u8def\u53e3\u3001\u5373\u5c07\u51fa\u73fe\u7684\u5c01\u9589\u8eca\u9053\u6216\u81ea\u884c\u8eca\u9053\u4e0a\u7684\u4f75\u6392\u505c\u8eca\u7b49\u8907\u96dc\u5834\u666f\u6642\uff0c\u7d93\u5e38\u9762\u81e8\u56f0\u5883\u3002\u63a8\u7406\u6280\u8853\u8ce6\u4e88\u81ea\u99d5\u8eca\u985e\u4f3c\u65bc\u4eba\u985e\u99d5\u99db\u7684\u5e38\u8b58\u5224\u65b7\u80fd\u529b\u3002<\/p>\n\n\n\n AR1 \u900f\u904e\u5206\u89e3\u5834\u666f\u4e26\u9010\u6b65\u9032\u884c\u63a8\u7406\u4f86\u5be6\u73fe\u9019\u4e00\u9ede\uff0c\u900f\u904e\u8a55\u4f30\u6240\u6709\u5834\u666f\u767c\u5c55\u53ef\u80fd\uff0c\u518d\u904b\u7528\u60c5\u5883\u8cc7\u6599\u9078\u64c7\u6700\u4f73\u8def\u7dda\u3002<\/p>\n\n\n\n \u8209\u4f8b\u800c\u8a00\uff0c\u900f\u904e AR1 \u5be6\u73fe\u7684\u601d\u7dad\u93c8\u63a8\u7406\uff0c\u81ea\u99d5\u8eca\u884c\u7d93\u884c\u4eba\u5bc6\u96c6\u5340\u4e14\u9130\u8fd1\u81ea\u884c\u8eca\u9053\u6642\uff0c\u8eca\u8f1b\u5f97\u4ee5\u5373\u6642\u6574\u5408\u8def\u5f91\u8cc7\u6599\u3001\u7d0d\u5165\u63a8\u7406\u8ecc\u8de1\u4f5c\u70ba\u63a1\u53d6\u7279\u5b9a\u884c\u52d5\u7684\u6c7a\u7b56\u4f9d\u64da\uff0c\u9032\u800c\u898f\u5283\u5f8c\u7e8c\u8def\u5f91\uff0c\u4f8b\u5982\u9060\u96e2\u81ea\u884c\u8eca\u9053\u6216\u70ba\u53ef\u80fd\u9055\u898f\u7a7f\u8d8a\u99ac\u8def\u7684\u884c\u4eba\u9810\u4f5c\u505c\u8eca\u6e96\u5099\u3002<\/p>\n\n\n\n \u57fa\u65bc NVIDIA Cosmos Reason<\/a> \u7684 AR1 \u958b\u653e\u578b\u57fa\u790e\uff0c\u8b93\u7814\u7a76\u4eba\u54e1\u53ef\u4ee5\u6839\u64da\u81ea\u8eab\u7684\u975e\u5546\u696d\u7528\u9014\u5ba2\u88fd\u5316\u6a21\u578b\uff0c\u7121\u8ad6\u662f\u7528\u65bc\u57fa\u6e96\u6e2c\u8a66\uff0c\u6216\u662f\u5efa\u7acb\u5be6\u9a57\u6027\u7684\u81ea\u52d5\u99d5\u99db\u61c9\u7528\u3002<\/p>\n\n\n\n \u5c0d\u65bc\u5f8c\u8a13\u7df4\u7684 AR1\uff0c\u5f37\u5316\u5b78\u7fd2<\/a>\u5df2\u88ab\u8b49\u660e\u7279\u5225\u6709\u6548\u3002\u7814\u7a76\u4eba\u54e1\u89c0\u5bdf\u5230\uff0c\u8207\u9810\u8a13\u7df4\u6a21\u578b\u76f8\u6bd4\uff0cAR1 \u7684\u63a8\u7406\u80fd\u529b\u986f\u8457\u63d0\u5347\u3002<\/p>\n\n\n\n NVIDIA DRIVE Alpamayo-R1 \u6a21\u578b\u5c07\u767c\u5e03\u65bc GitHub \u8207 Hugging Face\uff0c\u5176\u7528\u65bc\u8a13\u7df4\u548c\u8a55\u4f30\u8a72\u6a21\u578b\u7684\u8cc7\u6599\u5b50\u96c6\u4ea6\u767c\u5e03\u65bc NVIDIA Physical AI Open Dataset<\/a>\u3002NVIDIA \u4e5f\u767c\u5e03\u958b\u6e90\u7684 AlpaSim\u6846\u67b6<\/a>\uff0c\u7528\u65bc\u8a55\u4f30 AR1\u3002<\/p>\n\n\n\n \u4e86\u89e3\u66f4\u591a\u7528\u65bc\u81ea\u52d5\u99d5\u99db\u7684\u63a8\u7406\u8996\u89ba\u8a9e\u8a00\u52d5\u4f5c\u6a21\u578b<\/a>\u76f8\u95dc\u8cc7\u8a0a\u3002<\/p>\n\n\n\n <\/p>\n\n\n\n \u958b\u767c\u4eba\u54e1\u53ef\u4ee5\u900f\u904e Cosmos Cookbook<\/a> \u4e2d\u63d0\u4f9b\u7684\u9010\u6b65\u6559\u5b78\u3001\u5feb\u901f\u5165\u9580\u63a8\u8ad6\u7bc4\u4f8b\u548c\u9032\u968e\u5f8c\u8a13\u7df4\u5de5\u4f5c\u6d41\u7a0b\uff0c\u5b78\u7fd2\u5982\u4f55\u4f7f\u7528\u57fa\u65bc Cosmos \u7684\u6a21\u578b\u4e26\u9032\u884c\u5f8c\u8a13\u7df4\u3002\u9019\u4efd\u6307\u5357\u5c08\u70ba\u7269\u7406 AI \u958b\u767c\u4eba\u54e1\u6253\u9020\uff0c\u6db5\u84cb AI \u958b\u767c\u7684\u6bcf\u500b\u74b0\u7bc0\uff0c\u5305\u62ec\u8cc7\u6599\u6574\u7406\u3001\u5408\u6210\u8cc7\u6599\u751f\u6210<\/a>\u8207\u6a21\u578b\u8a55\u4f30\u3002<\/p>\n\n\n\n \u57fa\u65bc Cosmos \u7684\u61c9\u7528\u5c07\u64c1\u6709\u7121\u9650\u7684\u53ef\u80fd\u6027\u3002NVIDIA \u7684\u6700\u65b0\u61c9\u7528\u7bc4\u4f8b\u5305\u62ec\uff1a<\/p>\n\n\n\n \u3000\u25cf\u00a0\u00a0\u00a0LidarGen<\/a><\/strong>\uff0c\u5168\u7403\u9996\u500b\u80fd\u5920\u70ba\u81ea\u52d5\u99d5\u99db\uff08AV\uff09\u6a21\u64ec\u7522\u751f\u5149\u5b78\u96f7\u9054\u8cc7\u6599\u7684\u4e16\u754c\u6a21\u578b\u3002 <\/p>\n\n\n <\/p>\n\n\n\n \u7b56\u7565\u6a21\u578b\u53ef\u4ee5\u5728 NVIDIA Isaac Lab<\/a> \u8207 Isaac Sim<\/a> \u4e2d\u9032\u884c\u8a13\u7df4\uff0c\u800c\u5f9e\u7b56\u7565\u6a21\u578b\u751f\u6210\u7684\u8cc7\u6599\u80fd\u5920\u9032\u4e00\u6b65\u7528\u65bc\u5c0d NVIDIA GR00T N<\/a> \u6a5f\u5668\u4eba\u6a21\u578b\u9032\u884c\u5f8c\u8a13\u7df4\u3002<\/p>\n\n\n <\/p>\n\n\n\n NVIDIA \u751f\u614b\u7cfb\u5408\u4f5c\u5925\u4f34\u6b63\u904b\u7528 Cosmos \u4e16\u754c\u57fa\u790e\u6a21\u578b\u958b\u767c\u6700\u65b0\u6280\u8853\u3002<\/p>\n\n\n\n \u81ea\u99d5\u8eca\u958b\u767c\u5546 Voxel51<\/a> \u6b63\u70ba Cosmos Cookbook \u8ca2\u737b\u6a21\u578b\u914d\u65b9\u3002\u7269\u7406 AI \u958b\u767c\u5546 1X<\/a>\u3001Figure AI\u3001Foretellix\u3001Gatik\u3001Oxa\u3001PlusAI \u548c X-Humanoid<\/a> \u7686\u904b\u7528\u4e16\u754c\u57fa\u790e\u6a21\u578b\u958b\u767c\u6700\u65b0\u7269\u7406 AI \u61c9\u7528\u3002\u6b64\u5916\uff0c\u8607\u9ece\u4e16\u806f\u90a6\u7406\u5de5\u5b78\u9662\uff08ETH Zurich\uff09\u7684\u7814\u7a76\u4eba\u54e1\u4e5f\u5c07\u5728NeurIPS\u5927\u6703\u4e0a\u767c\u8868\u4e00\u7bc7\u8ad6\u6587<\/a>\uff0c\u91cd\u9ede\u4ecb\u7d39\u5982\u4f55\u4f7f\u7528 Cosmos \u6a21\u578b\u5275\u5efa\u903c\u771f\u4e14\u9023\u8cab\u7684 3D \u5834\u666f\u3002<\/p>\n\n\n\n <\/p>\n\n\n\n NVIDIA\u00a0\u540c\u6642\u63a8\u51fa\u591a\u91cd\u767c\u8a00\u8005\u8a9e\u97f3AI<\/a>\u6a21\u578b\u3001\u5177\u5099\u63a8\u7406\u80fd\u529b\u53ca AI \u5b89\u5168\u8cc7\u6599\u96c6\u7684\u65b0\u6a21\u578b\uff0c\u4ee5\u53ca\u7528\u65bc\u751f\u6210\u9ad8\u54c1\u8cea\u5408\u6210\u8cc7\u6599\u96c6\uff0c\u4f86\u5f37\u5316\u5b78\u7fd2\u548c\u7279\u5b9a\u9818\u57df\u6a21\u578b\u5ba2\u88fd\u5316\u7684\u958b\u653e\u5de5\u5177\u3002\u9019\u4e9b\u5de5\u5177\u5305\u62ec\uff1a<\/p>\n\n\n\n \u3000\u25cf\u00a0\u00a0\u00a0MultiTalker Parakeet<\/a>\uff1a<\/strong>\u7528\u65bc\u4e32\u6d41\u97f3\u8a0a\u7684\u81ea\u52d5\u8a9e\u97f3\u8fa8\u8b58\u6a21\u578b\uff0c\u5373\u4f7f\u5728\u5c0d\u8a71\u91cd\u758a\u6216\u8a9e\u901f\u5f88\u5feb\u7684\u60c5\u6cc1\u4e0b\uff0c\u4e5f\u80fd\u7406\u89e3\u591a\u4f4d\u767c\u8a00\u8005\u3002<\/p>\n\n\n\n \u3000\u25cf\u00a0\u00a0\u00a0Sortformer<\/a>\uff1a<\/strong>\u80fd\u5920\u5373\u6642\u6e96\u78ba\u5730\u5340\u5206\u97f3\u8a0a\u4e32\u6d41\u4e2d\u591a\u500b\u8aaa\u8a71\u8005\uff0c\u4ea6\u7a31\u70ba\u81ea\u52d5\u5206\u6bb5\u6a19\u8a18\uff08diarization\uff09\u7684\u5148\u9032\u6a21\u578b\u3002<\/p>\n\n\n\n \u3000\u25cf\u00a0\u00a0\u00a0Nemotron Content Safety Reasoning<\/a>\uff1a<\/strong>\u57fa\u65bc\u63a8\u7406\u7684 AI \u5b89\u5168\u6a21\u578b\uff0c\u53ef\u5728\u8de8\u9818\u57df\u52d5\u614b\u57f7\u884c\u5ba2\u88fd\u5316\u7b56\u7565\u3002<\/p>\n\n\n\n \u3000\u25cf\u00a0\u00a0\u00a0Nemotron Content Safety Audio Dataset<\/a>\uff1a<\/strong>\u7528\u65bc\u8a13\u7df4\u6a21\u578b\u4ee5\u5075\u6e2c\u4e0d\u5b89\u5168\u97f3\u8a0a\u5167\u5bb9\u7684\u5408\u6210\u8cc7\u6599\u96c6\uff0c\u5be6\u73fe\u8de8\u6587\u5b57\u548c\u97f3\u8a0a\u6a21\u614b\u7684\u5b89\u5168\u9632\u8b77\u6a5f\u5236\u958b\u767c\u3002<\/p>\n\n\n\n \u3000\u25cf\u00a0\u00a0\u00a0NeMo Gym<\/a>\uff1a<\/strong>\u4e00\u500b\u958b\u6e90\u51fd\u5f0f\u5eab\uff0c\u53ef\u52a0\u901f\u4e26\u7c21\u5316\u70ba\u5927\u578b\u8a9e\u8a00\u6a21\u578b\u8a13\u7df4\u958b\u767c\u7684\u5f37\u5316\u5b78\u7fd2\u74b0\u5883\u3002NeMo Gym \u540c\u6642\u6536\u9304\u65e5\u76ca\u64f4\u5c55\u7684\u5373\u7528\u578b\u8a13\u7df4\u74b0\u5883\u96c6\u5408\uff0c\u4ee5\u652f\u63f4\u57fa\u65bc\u53ef\u9a57\u8b49\u734e\u52f5\u7684\u5f37\u5316\u5b78\u7fd2\uff08RLVR\uff09\u3002<\/p>\n\n\n\n \u3000\u25cf\u00a0\u00a0\u00a0NeMo Data Designer Library<\/a>\uff1a<\/strong>\u8a72\u51fd\u5f0f\u5eab\u73fe\u5df2\u63a1\u7528 Apache 2.0 \u6388\u6b0a\u958b\u6e90\uff0c\u4e26\u63d0\u4f9b\u7aef\u5230\u7aef\u89e3\u6c7a\u65b9\u6848\uff0c\u7528\u65bc\u751f\u6210\u3001\u9a57\u8b49\u53ca\u6700\u4f73\u5316\u751f\u6210\u5f0f AI \u6240\u9700\u7684\u9ad8\u54c1\u8cea\u5408\u6210\u8cc7\u6599\u96c6\uff0c\u5305\u542b\u7279\u5b9a\u9818\u57df\u7684\u6a21\u578b\u5ba2\u88fd\u5316\u8207\u8a55\u4f30\u3002<\/p>\n\n\n\n CrowdStrike\u3001Palantir \u548c ServiceNow \u7b49 NVIDIA \u751f\u614b\u7cfb\u5408\u4f5c\u5925\u4f34\u6b63\u5728\u4f7f\u7528 NVIDIA Nemotron \u548c NeMo \u5de5\u5177\u5efa\u7acb\u5b89\u5168\u3001\u5c08\u696d\u7684\u4ee3\u7406\u578b AI\u3002<\/p>\n\n\n\n NeurIPS \u5927\u6703\u7684\u8207\u6703\u8005\u53ef\u65bc\u7f8e\u897f\u6642\u9593 12 \u6708 1 \u65e5\u4e0b\u5348 4 \u9ede\u81f3 8 \u9ede\u8209\u884c\u7684 Nemotron Summit<\/a> \u4e0a\u63a2\u7d22\u9019\u4e9b\u5275\u65b0\u6210\u679c\uff0cNVIDIA \u6df1\u5ea6\u5b78\u7fd2\u61c9\u7528\u7814\u7a76\u526f\u7e3d\u88c1 Bryan Catanzaro \u5c07\u767c\u8868\u958b\u5e55\u6f14\u8aaa\u3002<\/p>\n\n\n\n <\/p>\n\n\n\n NVIDIA\u5728NeurIPS\u5927\u6703\u4e0a\u767c\u8868\u4e86\u6578\u5341\u7bc7\u7814\u7a76\u8ad6\u6587<\/a>\uff0c\u4ee5\u4e0b\u70ba\u5e7e\u7bc7\u805a\u7126\u8a9e\u8a00\u6a21\u578b\u767c\u5c55\u7684\u4eae\u9ede\u6210\u679c\uff1a<\/p>\n\n\n\n \u3000\u25cf\u00a0\u00a0\u00a0Audio Flamingo 3: Advancing Audio Intelligence With Fully Open Large Audio Language Models<\/a>\uff1a<\/strong>\u9019\u6b3e\u5927\u578b\u97f3\u8a0a\u8a9e\u8a00\u6a21\u578b\u80fd\u5920\u5c0d\u8a9e\u97f3\u3001\u8072\u97f3\u548c\u97f3\u6a02\u9032\u884c\u63a8\u7406\u3002\u5b83\u53ef\u4ee5\u7406\u89e3\u548c\u63a8\u7406\u9577\u9054 10 \u5206\u9418\u7684\u97f3\u8a0a\u7247\u6bb5\uff0c\u4e26\u5728 20 \u591a\u500b\u57fa\u6e96\u6e2c\u8a66\u4e2d\u53d6\u5f97\u6700\u5148\u9032\u7684\u7d50\u679c\u3002 <\/p>\n\n\n\n \u67e5\u770bNeurIPS\u7684\u5b8c\u6574\u6d3b\u52d5\u5217\u8868<\/a>\uff0c\u5927\u6703\u5c07\u65bc\u7f8e\u570b\u8056\u5730\u7259\u54e5\u8209\u884c\u81f3 12 \u6708 7 \u65e5\uff08\u661f\u671f\u65e5\uff09\u3002<\/em><\/p>\n\n\n\n <\/p>\n\n\n\n \u3000\u3000\u3000<\/p>\n\n\n\n NVIDIA\u767c\u8868\u7528\u65bc\u8a9e\u97f3\u3001\u5b89\u5168\u8207\u81ea\u52d5<\/p>\n","protected":false},"author":3,"featured_media":131233,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"rank_math_lock_modified_date":false,"footnotes":""},"categories":[13],"tags":[16245,580,3548,71,14551],"class_list":["post-131230","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-news","tag-alpamayo-r1","tag-focus","tag-neurips","tag-nvidia","tag-ai"],"_links":{"self":[{"href":"https:\/\/www.ioiotimes.com\/index.php?rest_route=\/wp\/v2\/posts\/131230"}],"collection":[{"href":"https:\/\/www.ioiotimes.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.ioiotimes.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.ioiotimes.com\/index.php?rest_route=\/wp\/v2\/users\/3"}],"replies":[{"embeddable":true,"href":"https:\/\/www.ioiotimes.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=131230"}],"version-history":[{"count":1,"href":"https:\/\/www.ioiotimes.com\/index.php?rest_route=\/wp\/v2\/posts\/131230\/revisions"}],"predecessor-version":[{"id":131234,"href":"https:\/\/www.ioiotimes.com\/index.php?rest_route=\/wp\/v2\/posts\/131230\/revisions\/131234"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.ioiotimes.com\/index.php?rest_route=\/wp\/v2\/media\/131233"}],"wp:attachment":[{"href":"https:\/\/www.ioiotimes.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=131230"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.ioiotimes.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=131230"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.ioiotimes.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=131230"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}![\"\"](\"https:\/\/www.ioiotimes.com\/wordpress\/wp-content\/uploads\/2025\/12\/20251202-news-nvidia02.jpg\" "\\"\\"")
<\/figure>\n\n\n\nNVIDIA DRIVE Alpamayo-R1\u958b\u555f\u81ea\u52d5\u99d5\u99db\u7814\u7a76\u65b0\u7bc7\u7ae0<\/strong><\/h4>\n\n\n\n
\u70ba\u4efb\u4f55\u7269\u7406<\/strong>AI<\/strong>\u61c9\u7528\u5834\u666f\u5ba2\u88fd\u5316<\/strong>NVIDIA Cosmos<\/strong><\/h4>\n\n\n\n
\u3000\u25cf\u00a0\u00a0\u00a0Omniverse NuRec Fixer<\/a><\/strong>\uff0c\u7528\u65bc\u81ea\u52d5\u99d5\u99db\u548c\u6a5f\u5668\u4eba\u6a21\u64ec\u7684\u6a21\u578b\uff0c\u8a72\u6a21\u578b\u900f\u904e NVIDIA Cosmos<\/a>\u00a0Predict \u8fd1\u4e4e\u5373\u6642\u5730\u4fee\u6b63\u795e\u7d93\u91cd\u5efa\u8cc7\u6599\u4e2d\u7684\u7455\u75b5\uff0c\u4f8b\u5982\u4f86\u81ea\u65b0\u8996\u89d2\u6216\u96dc\u8a0a\u8cc7\u6599\u7684\u6a21\u7cca\u548c\u7f3a\u640d\u3002
\u3000\u25cf\u00a0\u00a0\u00a0Cosmos Policy<\/a><\/strong>\uff0c\u4e00\u500b\u5c07\u5927\u578b\u9810\u8a13\u7df4\u5f71\u7247\u6a21\u578b\u8f49\u63db\u70ba\u5065\u5168\u6a5f\u5668\u4eba\u7b56\u7565\uff08policy\uff09\u7684\u6846\u67b6\u3002\u6b64\u7b56\u7565\u662f\u4e00\u7d44\u7528\u65bc\u898f\u7bc4\u6a5f\u5668\u4eba\u884c\u70ba\u7684\u898f\u5247\u3002
\u3000\u25cf\u00a0\u00a0\u00a0ProtoMotions3<\/a><\/strong>\uff0c\u57fa\u65bc NVIDIA Newton \u548c Isaac Lab \u6253\u9020\u7684\u958b\u6e90\u3001GPU \u52a0\u901f\u6846\u67b6\uff0c\u7528\u65bc\u8a13\u7df4\u7269\u7406\u6a21\u64ec\u7684\u6578\u4f4d\u4eba\u985e\u548c\u4eba\u5f62\u6a5f\u5668\u4eba\uff0c\u5176\u8a13\u7df4\u7684\u903c\u771f\u5834\u666f\u7531 Cosmos \u4e16\u754c\u57fa\u790e\u6a21\u578b\uff08WFM\uff09<\/a>\u751f\u6210\u3002<\/p>\n\n\n\n![\"\"](\"https:\/\/www.ioiotimes.com\/wordpress\/wp-content\/uploads\/2025\/12\/20251202-news-nvidia03.jpg\" "\\"\\"")
![\"\"](\"https:\/\/www.ioiotimes.com\/wordpress\/wp-content\/uploads\/2025\/12\/20251202-news-nvidia04.gif\" "\\"\\"")
NVIDIA Nemotron<\/strong>\u65b0\u589e\u529f\u80fd\u5f37\u5316\u6578\u4f4d<\/strong>AI<\/strong>\u958b\u767c\u4eba\u54e1\u5de5\u5177\u7d44<\/strong><\/strong><\/h4>\n\n\n\n
NVIDIA<\/strong>\u7814\u7a76\u5718\u968a\u63a8\u52d5\u8a9e\u8a00<\/strong>AI<\/strong>\u5275\u65b0<\/strong><\/strong><\/h4>\n\n\n\n
\u3000\u25cf\u00a0\u00a0\u00a0Minitron-SSM: Efficient Hybrid Language Model Compression Through Group-Aware SSM Pruning<\/strong><\/a>\uff1a\u8a72\u7814\u7a76\u4ecb\u7d39\u4e00\u7a2e\u80fd\u58d3\u7e2e\u6df7\u5408\u6a21\u578b\u7684\u526a\u679d\u65b9\u6cd5\uff0c\u900f\u904e\u526a\u679d\u8207\u84b8\u993e\u5c07 Nemotron-H 8B<\/a> \u6a21\u578b\u53c3\u6578\u7531 80 \u5104\u58d3\u7e2e\u81f3 40 \u5104\u3002\u6700\u7d42\u7522\u51fa\u6a21\u578b\u7684\u6e96\u78ba\u7387\u4e0d\u50c5\u8d85\u8d8a\u540c\u7b49\u898f\u6a21\u7684\u6a21\u578b\uff0c\u540c\u6642\u63a8\u7406\u8f38\u9001\u91cf\u901f\u5ea6\u4e5f\u63d0\u9ad8\u81f3 2 \u500d\u3002
\u3000\u25cf\u00a0\u00a0\u00a0Jet-Nemotron: Efficient Language Model With Post Neural Architecture Search<\/a>\uff1a<\/strong>\u9019\u9805\u7814\u7a76\u63d0\u51fa\u4e00\u5957\u9ad8\u6210\u672c\u6548\u76ca\u7684\u5f8c\u8a13\u7df4\u6d41\u7a0b\uff0c\u7528\u65bc\u958b\u767c\u65b0\u578b\u9ad8\u6548\u8a9e\u8a00\u6a21\u578b\u67b6\u69cb\uff0c\u4e26\u4ecb\u7d39\u5229\u7528\u8a72\u6d41\u7a0b\u751f\u6210\u7684\u6df7\u5408\u67b6\u69cb\u6a21\u578b\u7cfb\u5217\u3002\u9019\u4e9b\u6a21\u578b\u5728\u5be6\u73fe\u751f\u6210\u8f38\u9001\u91cf\u986f\u8457\u63d0\u9ad8\u7684\u540c\u6642\uff0c\u5176\u6e96\u78ba\u7387\u8207\u9802\u5c16\u7684\u5168\u6ce8\u610f\u529b\u57fa\u7dda\uff08full-attention baseline\uff09\u6a21\u578b\u76f8\u7576\uff0c\u751a\u81f3\u66f4\u9ad8\u3002
\u3000\u25cf\u00a0\u00a0\u00a0Nemotron-Flash: Towards Latency-Optimal Hybrid Small Language Models<\/a>\uff1a<\/strong>\u672c\u5c08\u6848\u63d0\u51fa\u5168\u65b0\u7684\u5c0f\u578b\u8a9e\u8a00\u6a21\u578b\uff08SLM\uff09\u67b6\u69cb\uff0c\u4ee5\u5be6\u969b\u61c9\u7528\u4e2d\u7684\u5ef6\u9072\u800c\u975e\u53c3\u6578\u6578\u91cf\u70ba\u6838\u5fc3\u9032\u884c\u91cd\u65b0\u8a2d\u8a08\uff0c\u5be6\u73fe\u9802\u5c16\u7684\u901f\u5ea6\u8207\u6e96\u78ba\u7387\u3002
\u3000\u25cf\u00a0\u00a0\u00a0ProRL: Prolonged Reinforcement Learning Expands Reasoning Boundaries in Large Language Models<\/a><\/strong>\uff1a\u9577\u6642\u9593\u5f37\u5316\u5b78\u7fd2\uff08ProRL\uff09\u662f\u5ef6\u9577\u6a21\u578b\u8a13\u7df4\u6642\u9593\u5468\u671f\u7684\u6280\u8853\u3002\u5728 NeurIPS \u5927\u6703\u767c\u8868\u7684\u7814\u7a76\u6210\u679c\u4e2d\uff0cNVIDIA \u7814\u7a76\u4eba\u54e1\u95e1\u8ff0\u6b64\u65b9\u6cd5\u5982\u4f55\u4f7f\u6a21\u578b\u5728\u63a8\u7406\u80fd\u529b\u4e0a\u59cb\u7d42\u8d85\u8d8a\u57fa\u790e\u6a21\u578b\u3002<\/p>\n\n\n\n\ud83d\udfe6\u4ec0\u9ebc!?\u804a\u5929\u6253\u5c41\u4e5f\u53ef\u4ee5\u62bd\u734e\u52f5!! 11\u6708\u641e\u754f\u6709\u79ae\u6d3b\u52d5<\/a>\u73fe\u6b63\u9032\u884c\u4e2d!!
\ud83d\udfe6\u73fe\u5728\u5c31\u52a0\u5165 ioioTIMES \u81c9\u66f8\u7c89\u7d72\u5718 \u66f4\u591a\u4e92\u52d5\u3001\u66f4\u591a\u597d\u5eb7\u650f\u62b5\u52a0!!
\ud83d\udfe6\u6211\u5011\u6709LINE TODAY<\/a>\u983b\u9053\u4e86\uff0c\u5feb\u4f86\u8ffd\u8e2a\u6211\u5011\u5427!!–\u6700\u65b0\u79d1\u6280\u65b0\u805e \u76e1\u5728\u4f60<\/strong>\u624b<\/h4>\n","protected":false},"excerpt":{"rendered":"