routing_eval/validation.jsonl

{"id":"route_eval_dinov3_heldout","behavior":"ground_external","language":"en","context_keys":[],"evidence":{"entity":"DINOv3","definition":"DINOv3 is Meta's self-supervised vision-model family for learning general-purpose image representations and dense visual features without human labels.","sources":["https://ai.meta.com/research/dinov3/"]},"prompt":"Using the supplied evidence, what is DINOv3?","expected_groups":[["self-supervised"],["vision"],["without human labels","dense visual features"]],"forbidden_terms":["Daniel's work","few-shot","pose estimation"]}
{"id":"route_eval_deta_heldout","behavior":"ground_external","language":"en","context_keys":[],"evidence":{"entity":"DETA","definition":"DETA is an object detector from the NMS Strikes Back study that trains Deformable DETR with traditional one-to-many IoU-based label assignment and non-maximum suppression.","sources":["https://arxiv.org/abs/2212.06137"]},"prompt":"What does DETA mean in object detection?","expected_groups":[["object detector","object detection"],["Deformable DETR"],["one-to-many","IoU-based"]],"forbidden_terms":["Daniel created","few-shot learning"]}
{"id":"route_eval_unseen_evidence","behavior":"ground_external","language":"en","context_keys":[],"evidence":{"entity":"Contrastive learning","definition":"Contrastive learning learns representations by drawing related examples together and pushing dissimilar examples apart in an embedding space.","sources":["retrieved public source"]},"prompt":"Explain contrastive learning from the evidence in one sentence.","expected_groups":[["representations"],["related","similar"],["dissimilar","apart"]],"forbidden_terms":["Daniel","portfolio"]}
{"id":"route_eval_uiuc_definition","behavior":"ground_external","language":"en","context_keys":[],"evidence":{"entity":"UIUC","definition":"UIUC is the University of Illinois Urbana-Champaign, located in the Champaign-Urbana community in Illinois, United States.","sources":["https://www.admissions.illinois.edu/about"]},"prompt":"I am asking about the institution itself: where is UIUC?","expected_groups":[["University of Illinois Urbana-Champaign"],["Champaign-Urbana"],["Illinois"]],"forbidden_terms":["KAIST","POSTECH","exchange student"]}
{"id":"route_eval_rt_relation","behavior":"answer","language":"en","context_keys":["open_source"],"evidence":{"entity":"RT-DETR","profile_relation":"Daniel contributed RT-DETR support to Hugging Face Transformers, including model implementation and training-related integration work."},"prompt":"I know what RT-DETR is. What was Daniel's contribution to it?","expected_groups":[["RT-DETR"],["Hugging Face Transformers","Transformers"],["training","implementation"]],"forbidden_terms":["negative sampling","few-shot learning","Daniel created RT-DETR"]}
{"id":"route_eval_vitpose_definition","behavior":"ground_external","language":"en","context_keys":[],"evidence":{"entity":"ViTPose","definition":"ViTPose is a human-pose-estimation model that uses a plain Vision Transformer backbone and a lightweight decoder to predict body keypoints.","sources":["https://arxiv.org/abs/2204.12484"]},"prompt":"Describe ViTPose itself rather than Daniel's contribution.","expected_groups":[["pose"],["Vision Transformer"],["keypoints","decoder"]],"forbidden_terms":["few-shot learning","position and orientation","Daniel's work"]}
{"id":"route_eval_retrieve_cmu","behavior":"retrieve","language":"en","context_keys":[],"prompt":"Where is Carnegie Mellon University?","expected_groups":[["<search_public_knowledge>"],["Carnegie Mellon University"],["</search_public_knowledge>"]],"forbidden_terms":["Daniel","KAIST","Pittsburgh"]}
{"id":"route_eval_retrieve_clip","behavior":"retrieve","language":"en","context_keys":[],"prompt":"What is CLIP in machine learning?","expected_groups":[["<search_public_knowledge>"],["CLIP"],["</search_public_knowledge>"]],"forbidden_terms":["Daniel's","few-shot"]}
{"id":"route_eval_retrieve_ko","behavior":"retrieve","language":"ko","context_keys":[],"prompt":"NeRF가 무엇인지 알려줘.","expected_groups":[["<search_public_knowledge>"],["NeRF"],["</search_public_knowledge>"]],"forbidden_terms":["Daniel","포트폴리오"]}