Add files using upload-large-folder tool

q11/control_k0/question.json

@@ -0,0 +1,25 @@
+{
+  "cell_id": "q11_control_k0",
+  "doc_id": "2310.05634v2.pdf",
+  "question": "According to author's definition on conscious incompetence, when can a sentence map to both [NA] and a list of sub-graph knowledge?",
+  "answer": "If it can be partially verified by the knowledge graph G",
+  "answer_format": "Str",
+  "condition": "control",
+  "k": 0,
+  "source_position": "first",
+  "pdf_paths": [
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/pdfs/2310.05634v2.pdf"
+  ],
+  "pdf_names": [
+    "2310.05634v2.pdf"
+  ],
+  "n_pages": 24,
+  "est_tokens": 9600,
+  "expected_route": "azure",
+  "bundle_filenames": [
+    "source.pdf"
+  ],
+  "original_filenames": [
+    "2310.05634v2.pdf"
+  ]
+}

q11/hard_negative_k2/question.json

@@ -0,0 +1,33 @@
+{
+  "cell_id": "q11_hard_negative_k2",
+  "doc_id": "2310.05634v2.pdf",
+  "question": "According to author's definition on conscious incompetence, when can a sentence map to both [NA] and a list of sub-graph knowledge?",
+  "answer": "If it can be partially verified by the knowledge graph G",
+  "answer_format": "Str",
+  "condition": "hard_negative",
+  "k": 2,
+  "source_position": "first",
+  "pdf_paths": [
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/pdfs/2310.05634v2.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/5349694f4eecdd46092594856d170286c4d2fc94.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_a5b939c47368269e.pdf"
+  ],
+  "pdf_names": [
+    "2310.05634v2.pdf",
+    "5349694f4eecdd46092594856d170286c4d2fc94.pdf",
+    "web_a5b939c47368269e.pdf"
+  ],
+  "n_pages": 75,
+  "est_tokens": 30000,
+  "expected_route": "openrouter",
+  "bundle_filenames": [
+    "source.pdf",
+    "hn_1.pdf",
+    "hn_2.pdf"
+  ],
+  "original_filenames": [
+    "2310.05634v2.pdf",
+    "5349694f4eecdd46092594856d170286c4d2fc94.pdf",
+    "web_a5b939c47368269e.pdf"
+  ]
+}

q11/hard_negative_k4/question.json

@@ -0,0 +1,41 @@
+{
+  "cell_id": "q11_hard_negative_k4",
+  "doc_id": "2310.05634v2.pdf",
+  "question": "According to author's definition on conscious incompetence, when can a sentence map to both [NA] and a list of sub-graph knowledge?",
+  "answer": "If it can be partially verified by the knowledge graph G",
+  "answer_format": "Str",
+  "condition": "hard_negative",
+  "k": 4,
+  "source_position": "first",
+  "pdf_paths": [
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/pdfs/2310.05634v2.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/5349694f4eecdd46092594856d170286c4d2fc94.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_a5b939c47368269e.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_ddfd3d214b1f0ea4.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/1d1e9796cf93063d60a92e083ff4372e3ac27d28.pdf"
+  ],
+  "pdf_names": [
+    "2310.05634v2.pdf",
+    "5349694f4eecdd46092594856d170286c4d2fc94.pdf",
+    "web_a5b939c47368269e.pdf",
+    "web_ddfd3d214b1f0ea4.pdf",
+    "1d1e9796cf93063d60a92e083ff4372e3ac27d28.pdf"
+  ],
+  "n_pages": 118,
+  "est_tokens": 47200,
+  "expected_route": "openrouter",
+  "bundle_filenames": [
+    "source.pdf",
+    "hn_1.pdf",
+    "hn_2.pdf",
+    "hn_3.pdf",
+    "hn_4.pdf"
+  ],
+  "original_filenames": [
+    "2310.05634v2.pdf",
+    "5349694f4eecdd46092594856d170286c4d2fc94.pdf",
+    "web_a5b939c47368269e.pdf",
+    "web_ddfd3d214b1f0ea4.pdf",
+    "1d1e9796cf93063d60a92e083ff4372e3ac27d28.pdf"
+  ]
+}

q11/random_k2/question.json

@@ -0,0 +1,33 @@
+{
+  "cell_id": "q11_random_k2",
+  "doc_id": "2310.05634v2.pdf",
+  "question": "According to author's definition on conscious incompetence, when can a sentence map to both [NA] and a list of sub-graph knowledge?",
+  "answer": "If it can be partially verified by the knowledge graph G",
+  "answer_format": "Str",
+  "condition": "random",
+  "k": 2,
+  "source_position": "first",
+  "pdf_paths": [
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/pdfs/2310.05634v2.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_614534f8a63c89ee.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/39b8fca5d6df621aefdf7b7c61d925b79423ca25.pdf"
+  ],
+  "pdf_names": [
+    "2310.05634v2.pdf",
+    "web_614534f8a63c89ee.pdf",
+    "39b8fca5d6df621aefdf7b7c61d925b79423ca25.pdf"
+  ],
+  "n_pages": 63,
+  "est_tokens": 25200,
+  "expected_route": "openrouter",
+  "bundle_filenames": [
+    "source.pdf",
+    "random_1.pdf",
+    "random_2.pdf"
+  ],
+  "original_filenames": [
+    "2310.05634v2.pdf",
+    "web_614534f8a63c89ee.pdf",
+    "39b8fca5d6df621aefdf7b7c61d925b79423ca25.pdf"
+  ]
+}

q11/random_k4/question.json

@@ -0,0 +1,41 @@
+{
+  "cell_id": "q11_random_k4",
+  "doc_id": "2310.05634v2.pdf",
+  "question": "According to author's definition on conscious incompetence, when can a sentence map to both [NA] and a list of sub-graph knowledge?",
+  "answer": "If it can be partially verified by the knowledge graph G",
+  "answer_format": "Str",
+  "condition": "random",
+  "k": 4,
+  "source_position": "first",
+  "pdf_paths": [
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/pdfs/2310.05634v2.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_614534f8a63c89ee.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/39b8fca5d6df621aefdf7b7c61d925b79423ca25.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_38335edb03fd4d69.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_db5391cbc08bcba6.pdf"
+  ],
+  "pdf_names": [
+    "2310.05634v2.pdf",
+    "web_614534f8a63c89ee.pdf",
+    "39b8fca5d6df621aefdf7b7c61d925b79423ca25.pdf",
+    "web_38335edb03fd4d69.pdf",
+    "web_db5391cbc08bcba6.pdf"
+  ],
+  "n_pages": 281,
+  "est_tokens": 112400,
+  "expected_route": "openrouter",
+  "bundle_filenames": [
+    "source.pdf",
+    "random_1.pdf",
+    "random_2.pdf",
+    "random_3.pdf",
+    "random_4.pdf"
+  ],
+  "original_filenames": [
+    "2310.05634v2.pdf",
+    "web_614534f8a63c89ee.pdf",
+    "39b8fca5d6df621aefdf7b7c61d925b79423ca25.pdf",
+    "web_38335edb03fd4d69.pdf",
+    "web_db5391cbc08bcba6.pdf"
+  ]
+}

q16/control_k0/question.json

@@ -0,0 +1,25 @@
+{
+  "cell_id": "q16_control_k0",
+  "doc_id": "2312.10997v5.pdf",
+  "question": "According to table II, which are the datasets that has exactly three methods?",
+  "answer": "['POPQA', 'MS MARCO', 'SST-2']",
+  "answer_format": "List",
+  "condition": "control",
+  "k": 0,
+  "source_position": "first",
+  "pdf_paths": [
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/pdfs/2312.10997v5.pdf"
+  ],
+  "pdf_names": [
+    "2312.10997v5.pdf"
+  ],
+  "n_pages": 21,
+  "est_tokens": 8400,
+  "expected_route": "azure",
+  "bundle_filenames": [
+    "source.pdf"
+  ],
+  "original_filenames": [
+    "2312.10997v5.pdf"
+  ]
+}

q16/hard_negative_k2/question.json

@@ -0,0 +1,33 @@
+{
+  "cell_id": "q16_hard_negative_k2",
+  "doc_id": "2312.10997v5.pdf",
+  "question": "According to table II, which are the datasets that has exactly three methods?",
+  "answer": "['POPQA', 'MS MARCO', 'SST-2']",
+  "answer_format": "List",
+  "condition": "hard_negative",
+  "k": 2,
+  "source_position": "first",
+  "pdf_paths": [
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/pdfs/2312.10997v5.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_54e5ed38ff6e9c8d.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/a8431481e88e12d57e4fcf2c30c9d3977e16f1ad.pdf"
+  ],
+  "pdf_names": [
+    "2312.10997v5.pdf",
+    "web_54e5ed38ff6e9c8d.pdf",
+    "a8431481e88e12d57e4fcf2c30c9d3977e16f1ad.pdf"
+  ],
+  "n_pages": 61,
+  "est_tokens": 24400,
+  "expected_route": "openrouter",
+  "bundle_filenames": [
+    "source.pdf",
+    "hn_1.pdf",
+    "hn_2.pdf"
+  ],
+  "original_filenames": [
+    "2312.10997v5.pdf",
+    "web_54e5ed38ff6e9c8d.pdf",
+    "a8431481e88e12d57e4fcf2c30c9d3977e16f1ad.pdf"
+  ]
+}

q16/hard_negative_k4/question.json

@@ -0,0 +1,41 @@
+{
+  "cell_id": "q16_hard_negative_k4",
+  "doc_id": "2312.10997v5.pdf",
+  "question": "According to table II, which are the datasets that has exactly three methods?",
+  "answer": "['POPQA', 'MS MARCO', 'SST-2']",
+  "answer_format": "List",
+  "condition": "hard_negative",
+  "k": 4,
+  "source_position": "first",
+  "pdf_paths": [
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/pdfs/2312.10997v5.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_54e5ed38ff6e9c8d.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/a8431481e88e12d57e4fcf2c30c9d3977e16f1ad.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/3becf3cee2ec9db3b4bdae0c4504cf6821a6d8e6.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/dbd415315edbccc4954de733324058858328edb8.pdf"
+  ],
+  "pdf_names": [
+    "2312.10997v5.pdf",
+    "web_54e5ed38ff6e9c8d.pdf",
+    "a8431481e88e12d57e4fcf2c30c9d3977e16f1ad.pdf",
+    "3becf3cee2ec9db3b4bdae0c4504cf6821a6d8e6.pdf",
+    "dbd415315edbccc4954de733324058858328edb8.pdf"
+  ],
+  "n_pages": 74,
+  "est_tokens": 29600,
+  "expected_route": "openrouter",
+  "bundle_filenames": [
+    "source.pdf",
+    "hn_1.pdf",
+    "hn_2.pdf",
+    "hn_3.pdf",
+    "hn_4.pdf"
+  ],
+  "original_filenames": [
+    "2312.10997v5.pdf",
+    "web_54e5ed38ff6e9c8d.pdf",
+    "a8431481e88e12d57e4fcf2c30c9d3977e16f1ad.pdf",
+    "3becf3cee2ec9db3b4bdae0c4504cf6821a6d8e6.pdf",
+    "dbd415315edbccc4954de733324058858328edb8.pdf"
+  ]
+}

q16/random_k2/question.json

@@ -0,0 +1,33 @@
+{
+  "cell_id": "q16_random_k2",
+  "doc_id": "2312.10997v5.pdf",
+  "question": "According to table II, which are the datasets that has exactly three methods?",
+  "answer": "['POPQA', 'MS MARCO', 'SST-2']",
+  "answer_format": "List",
+  "condition": "random",
+  "k": 2,
+  "source_position": "first",
+  "pdf_paths": [
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/pdfs/2312.10997v5.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_9f3c2f30cbfcccca.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_f2077baa5993127a.pdf"
+  ],
+  "pdf_names": [
+    "2312.10997v5.pdf",
+    "web_9f3c2f30cbfcccca.pdf",
+    "web_f2077baa5993127a.pdf"
+  ],
+  "n_pages": 324,
+  "est_tokens": 129600,
+  "expected_route": "openrouter",
+  "bundle_filenames": [
+    "source.pdf",
+    "random_1.pdf",
+    "random_2.pdf"
+  ],
+  "original_filenames": [
+    "2312.10997v5.pdf",
+    "web_9f3c2f30cbfcccca.pdf",
+    "web_f2077baa5993127a.pdf"
+  ]
+}

q16/random_k4/question.json

@@ -0,0 +1,41 @@
+{
+  "cell_id": "q16_random_k4",
+  "doc_id": "2312.10997v5.pdf",
+  "question": "According to table II, which are the datasets that has exactly three methods?",
+  "answer": "['POPQA', 'MS MARCO', 'SST-2']",
+  "answer_format": "List",
+  "condition": "random",
+  "k": 4,
+  "source_position": "first",
+  "pdf_paths": [
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/pdfs/2312.10997v5.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_9f3c2f30cbfcccca.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_f2077baa5993127a.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_4766ecdcaaa8fd22.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_dc179c2563f3e3ab.pdf"
+  ],
+  "pdf_names": [
+    "2312.10997v5.pdf",
+    "web_9f3c2f30cbfcccca.pdf",
+    "web_f2077baa5993127a.pdf",
+    "web_4766ecdcaaa8fd22.pdf",
+    "web_dc179c2563f3e3ab.pdf"
+  ],
+  "n_pages": 430,
+  "est_tokens": 172000,
+  "expected_route": "openrouter",
+  "bundle_filenames": [
+    "source.pdf",
+    "random_1.pdf",
+    "random_2.pdf",
+    "random_3.pdf",
+    "random_4.pdf"
+  ],
+  "original_filenames": [
+    "2312.10997v5.pdf",
+    "web_9f3c2f30cbfcccca.pdf",
+    "web_f2077baa5993127a.pdf",
+    "web_4766ecdcaaa8fd22.pdf",
+    "web_dc179c2563f3e3ab.pdf"
+  ]
+}

q18/control_k0/question.json

@@ -0,0 +1,25 @@
+{
+  "cell_id": "q18_control_k0",
+  "doc_id": "2312.10997v5.pdf",
+  "question": "Which method integrates knowledge into white-box models via directive fine-tuning?",
+  "answer": "PKG",
+  "answer_format": "Str",
+  "condition": "control",
+  "k": 0,
+  "source_position": "first",
+  "pdf_paths": [
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/pdfs/2312.10997v5.pdf"
+  ],
+  "pdf_names": [
+    "2312.10997v5.pdf"
+  ],
+  "n_pages": 21,
+  "est_tokens": 8400,
+  "expected_route": "azure",
+  "bundle_filenames": [
+    "source.pdf"
+  ],
+  "original_filenames": [
+    "2312.10997v5.pdf"
+  ]
+}

q18/hard_negative_k2/question.json

@@ -0,0 +1,33 @@
+{
+  "cell_id": "q18_hard_negative_k2",
+  "doc_id": "2312.10997v5.pdf",
+  "question": "Which method integrates knowledge into white-box models via directive fine-tuning?",
+  "answer": "PKG",
+  "answer_format": "Str",
+  "condition": "hard_negative",
+  "k": 2,
+  "source_position": "first",
+  "pdf_paths": [
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/pdfs/2312.10997v5.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_54e5ed38ff6e9c8d.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/a8431481e88e12d57e4fcf2c30c9d3977e16f1ad.pdf"
+  ],
+  "pdf_names": [
+    "2312.10997v5.pdf",
+    "web_54e5ed38ff6e9c8d.pdf",
+    "a8431481e88e12d57e4fcf2c30c9d3977e16f1ad.pdf"
+  ],
+  "n_pages": 61,
+  "est_tokens": 24400,
+  "expected_route": "openrouter",
+  "bundle_filenames": [
+    "source.pdf",
+    "hn_1.pdf",
+    "hn_2.pdf"
+  ],
+  "original_filenames": [
+    "2312.10997v5.pdf",
+    "web_54e5ed38ff6e9c8d.pdf",
+    "a8431481e88e12d57e4fcf2c30c9d3977e16f1ad.pdf"
+  ]
+}

q18/hard_negative_k4/question.json

@@ -0,0 +1,41 @@
+{
+  "cell_id": "q18_hard_negative_k4",
+  "doc_id": "2312.10997v5.pdf",
+  "question": "Which method integrates knowledge into white-box models via directive fine-tuning?",
+  "answer": "PKG",
+  "answer_format": "Str",
+  "condition": "hard_negative",
+  "k": 4,
+  "source_position": "first",
+  "pdf_paths": [
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/pdfs/2312.10997v5.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_54e5ed38ff6e9c8d.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/a8431481e88e12d57e4fcf2c30c9d3977e16f1ad.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/3becf3cee2ec9db3b4bdae0c4504cf6821a6d8e6.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/dbd415315edbccc4954de733324058858328edb8.pdf"
+  ],
+  "pdf_names": [
+    "2312.10997v5.pdf",
+    "web_54e5ed38ff6e9c8d.pdf",
+    "a8431481e88e12d57e4fcf2c30c9d3977e16f1ad.pdf",
+    "3becf3cee2ec9db3b4bdae0c4504cf6821a6d8e6.pdf",
+    "dbd415315edbccc4954de733324058858328edb8.pdf"
+  ],
+  "n_pages": 74,
+  "est_tokens": 29600,
+  "expected_route": "openrouter",
+  "bundle_filenames": [
+    "source.pdf",
+    "hn_1.pdf",
+    "hn_2.pdf",
+    "hn_3.pdf",
+    "hn_4.pdf"
+  ],
+  "original_filenames": [
+    "2312.10997v5.pdf",
+    "web_54e5ed38ff6e9c8d.pdf",
+    "a8431481e88e12d57e4fcf2c30c9d3977e16f1ad.pdf",
+    "3becf3cee2ec9db3b4bdae0c4504cf6821a6d8e6.pdf",
+    "dbd415315edbccc4954de733324058858328edb8.pdf"
+  ]
+}

q18/random_k2/question.json

@@ -0,0 +1,33 @@
+{
+  "cell_id": "q18_random_k2",
+  "doc_id": "2312.10997v5.pdf",
+  "question": "Which method integrates knowledge into white-box models via directive fine-tuning?",
+  "answer": "PKG",
+  "answer_format": "Str",
+  "condition": "random",
+  "k": 2,
+  "source_position": "first",
+  "pdf_paths": [
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/pdfs/2312.10997v5.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_9f3c2f30cbfcccca.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_f2077baa5993127a.pdf"
+  ],
+  "pdf_names": [
+    "2312.10997v5.pdf",
+    "web_9f3c2f30cbfcccca.pdf",
+    "web_f2077baa5993127a.pdf"
+  ],
+  "n_pages": 324,
+  "est_tokens": 129600,
+  "expected_route": "openrouter",
+  "bundle_filenames": [
+    "source.pdf",
+    "random_1.pdf",
+    "random_2.pdf"
+  ],
+  "original_filenames": [
+    "2312.10997v5.pdf",
+    "web_9f3c2f30cbfcccca.pdf",
+    "web_f2077baa5993127a.pdf"
+  ]
+}

q18/random_k4/question.json

@@ -0,0 +1,41 @@
+{
+  "cell_id": "q18_random_k4",
+  "doc_id": "2312.10997v5.pdf",
+  "question": "Which method integrates knowledge into white-box models via directive fine-tuning?",
+  "answer": "PKG",
+  "answer_format": "Str",
+  "condition": "random",
+  "k": 4,
+  "source_position": "first",
+  "pdf_paths": [
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/pdfs/2312.10997v5.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_9f3c2f30cbfcccca.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_f2077baa5993127a.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_4766ecdcaaa8fd22.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_dc179c2563f3e3ab.pdf"
+  ],
+  "pdf_names": [
+    "2312.10997v5.pdf",
+    "web_9f3c2f30cbfcccca.pdf",
+    "web_f2077baa5993127a.pdf",
+    "web_4766ecdcaaa8fd22.pdf",
+    "web_dc179c2563f3e3ab.pdf"
+  ],
+  "n_pages": 430,
+  "est_tokens": 172000,
+  "expected_route": "openrouter",
+  "bundle_filenames": [
+    "source.pdf",
+    "random_1.pdf",
+    "random_2.pdf",
+    "random_3.pdf",
+    "random_4.pdf"
+  ],
+  "original_filenames": [
+    "2312.10997v5.pdf",
+    "web_9f3c2f30cbfcccca.pdf",
+    "web_f2077baa5993127a.pdf",
+    "web_4766ecdcaaa8fd22.pdf",
+    "web_dc179c2563f3e3ab.pdf"
+  ]
+}

q20/control_k0/question.json

@@ -0,0 +1,25 @@
+{
+  "cell_id": "q20_control_k0",
+  "doc_id": "nova_y70.pdf",
+  "question": "What devices other than phone are introduced for setting device+?",
+  "answer": "['vision', 'bluetooth device']",
+  "answer_format": "List",
+  "condition": "control",
+  "k": 0,
+  "source_position": "first",
+  "pdf_paths": [
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/pdfs/nova_y70.pdf"
+  ],
+  "pdf_names": [
+    "nova_y70.pdf"
+  ],
+  "n_pages": 45,
+  "est_tokens": 18000,
+  "expected_route": "azure",
+  "bundle_filenames": [
+    "source.pdf"
+  ],
+  "original_filenames": [
+    "nova_y70.pdf"
+  ]
+}

q20/hard_negative_k2/question.json

@@ -0,0 +1,33 @@
+{
+  "cell_id": "q20_hard_negative_k2",
+  "doc_id": "nova_y70.pdf",
+  "question": "What devices other than phone are introduced for setting device+?",
+  "answer": "['vision', 'bluetooth device']",
+  "answer_format": "List",
+  "condition": "hard_negative",
+  "k": 2,
+  "source_position": "first",
+  "pdf_paths": [
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/pdfs/nova_y70.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_14dc67230c1ac4c6.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_0b29bc1a42f15795.pdf"
+  ],
+  "pdf_names": [
+    "nova_y70.pdf",
+    "web_14dc67230c1ac4c6.pdf",
+    "web_0b29bc1a42f15795.pdf"
+  ],
+  "n_pages": 148,
+  "est_tokens": 59200,
+  "expected_route": "openrouter",
+  "bundle_filenames": [
+    "source.pdf",
+    "hn_1.pdf",
+    "hn_2.pdf"
+  ],
+  "original_filenames": [
+    "nova_y70.pdf",
+    "web_14dc67230c1ac4c6.pdf",
+    "web_0b29bc1a42f15795.pdf"
+  ]
+}

q20/hard_negative_k4/question.json

@@ -0,0 +1,41 @@
+{
+  "cell_id": "q20_hard_negative_k4",
+  "doc_id": "nova_y70.pdf",
+  "question": "What devices other than phone are introduced for setting device+?",
+  "answer": "['vision', 'bluetooth device']",
+  "answer_format": "List",
+  "condition": "hard_negative",
+  "k": 4,
+  "source_position": "first",
+  "pdf_paths": [
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/pdfs/nova_y70.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_14dc67230c1ac4c6.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_0b29bc1a42f15795.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_f6b882f6f2e37961.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_435e2f111b47ee17.pdf"
+  ],
+  "pdf_names": [
+    "nova_y70.pdf",
+    "web_14dc67230c1ac4c6.pdf",
+    "web_0b29bc1a42f15795.pdf",
+    "web_f6b882f6f2e37961.pdf",
+    "web_435e2f111b47ee17.pdf"
+  ],
+  "n_pages": 300,
+  "est_tokens": 120000,
+  "expected_route": "openrouter",
+  "bundle_filenames": [
+    "source.pdf",
+    "hn_1.pdf",
+    "hn_2.pdf",
+    "hn_3.pdf",
+    "hn_4.pdf"
+  ],
+  "original_filenames": [
+    "nova_y70.pdf",
+    "web_14dc67230c1ac4c6.pdf",
+    "web_0b29bc1a42f15795.pdf",
+    "web_f6b882f6f2e37961.pdf",
+    "web_435e2f111b47ee17.pdf"
+  ]
+}

q20/random_k2/question.json

@@ -0,0 +1,33 @@
+{
+  "cell_id": "q20_random_k2",
+  "doc_id": "nova_y70.pdf",
+  "question": "What devices other than phone are introduced for setting device+?",
+  "answer": "['vision', 'bluetooth device']",
+  "answer_format": "List",
+  "condition": "random",
+  "k": 2,
+  "source_position": "first",
+  "pdf_paths": [
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/pdfs/nova_y70.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_fe44495665edf65a.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_ee74e05dd5553065.pdf"
+  ],
+  "pdf_names": [
+    "nova_y70.pdf",
+    "web_fe44495665edf65a.pdf",
+    "web_ee74e05dd5553065.pdf"
+  ],
+  "n_pages": 214,
+  "est_tokens": 85600,
+  "expected_route": "openrouter",
+  "bundle_filenames": [
+    "source.pdf",
+    "random_1.pdf",
+    "random_2.pdf"
+  ],
+  "original_filenames": [
+    "nova_y70.pdf",
+    "web_fe44495665edf65a.pdf",
+    "web_ee74e05dd5553065.pdf"
+  ]
+}

q20/random_k4/question.json

@@ -0,0 +1,41 @@
+{
+  "cell_id": "q20_random_k4",
+  "doc_id": "nova_y70.pdf",
+  "question": "What devices other than phone are introduced for setting device+?",
+  "answer": "['vision', 'bluetooth device']",
+  "answer_format": "List",
+  "condition": "random",
+  "k": 4,
+  "source_position": "first",
+  "pdf_paths": [
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/pdfs/nova_y70.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_fe44495665edf65a.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_ee74e05dd5553065.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_9a06b1637bfaf79e.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_06f69d0a173b6b98.pdf"
+  ],
+  "pdf_names": [
+    "nova_y70.pdf",
+    "web_fe44495665edf65a.pdf",
+    "web_ee74e05dd5553065.pdf",
+    "web_9a06b1637bfaf79e.pdf",
+    "web_06f69d0a173b6b98.pdf"
+  ],
+  "n_pages": 375,
+  "est_tokens": 150000,
+  "expected_route": "openrouter",
+  "bundle_filenames": [
+    "source.pdf",
+    "random_1.pdf",
+    "random_2.pdf",
+    "random_3.pdf",
+    "random_4.pdf"
+  ],
+  "original_filenames": [
+    "nova_y70.pdf",
+    "web_fe44495665edf65a.pdf",
+    "web_ee74e05dd5553065.pdf",
+    "web_9a06b1637bfaf79e.pdf",
+    "web_06f69d0a173b6b98.pdf"
+  ]
+}

q203/control_k0/question.json

@@ -0,0 +1,25 @@
+{
+  "cell_id": "q203_control_k0",
+  "doc_id": "2021-Apple-Catalog.pdf",
+  "question": "One40 can only be used for Apple Watch, is that true? Please directly answer 'yes' or 'no'.",
+  "answer": "Yes",
+  "answer_format": "Str",
+  "condition": "control",
+  "k": 0,
+  "source_position": "first",
+  "pdf_paths": [
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/pdfs/2021-Apple-Catalog.pdf"
+  ],
+  "pdf_names": [
+    "2021-Apple-Catalog.pdf"
+  ],
+  "n_pages": 55,
+  "est_tokens": 22000,
+  "expected_route": "openrouter",
+  "bundle_filenames": [
+    "source.pdf"
+  ],
+  "original_filenames": [
+    "2021-Apple-Catalog.pdf"
+  ]
+}

q203/hard_negative_k2/question.json

@@ -0,0 +1,33 @@
+{
+  "cell_id": "q203_hard_negative_k2",
+  "doc_id": "2021-Apple-Catalog.pdf",
+  "question": "One40 can only be used for Apple Watch, is that true? Please directly answer 'yes' or 'no'.",
+  "answer": "Yes",
+  "answer_format": "Str",
+  "condition": "hard_negative",
+  "k": 2,
+  "source_position": "first",
+  "pdf_paths": [
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/pdfs/2021-Apple-Catalog.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_6820ef1c6bfc9e7b.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_935425021e4da339.pdf"
+  ],
+  "pdf_names": [
+    "2021-Apple-Catalog.pdf",
+    "web_6820ef1c6bfc9e7b.pdf",
+    "web_935425021e4da339.pdf"
+  ],
+  "n_pages": 140,
+  "est_tokens": 56000,
+  "expected_route": "openrouter",
+  "bundle_filenames": [
+    "source.pdf",
+    "hn_1.pdf",
+    "hn_2.pdf"
+  ],
+  "original_filenames": [
+    "2021-Apple-Catalog.pdf",
+    "web_6820ef1c6bfc9e7b.pdf",
+    "web_935425021e4da339.pdf"
+  ]
+}

q203/hard_negative_k4/question.json

@@ -0,0 +1,41 @@
+{
+  "cell_id": "q203_hard_negative_k4",
+  "doc_id": "2021-Apple-Catalog.pdf",
+  "question": "One40 can only be used for Apple Watch, is that true? Please directly answer 'yes' or 'no'.",
+  "answer": "Yes",
+  "answer_format": "Str",
+  "condition": "hard_negative",
+  "k": 4,
+  "source_position": "first",
+  "pdf_paths": [
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/pdfs/2021-Apple-Catalog.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_6820ef1c6bfc9e7b.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_935425021e4da339.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_d939e3897d0c1a86.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_449e1fe2966b96f2.pdf"
+  ],
+  "pdf_names": [
+    "2021-Apple-Catalog.pdf",
+    "web_6820ef1c6bfc9e7b.pdf",
+    "web_935425021e4da339.pdf",
+    "web_d939e3897d0c1a86.pdf",
+    "web_449e1fe2966b96f2.pdf"
+  ],
+  "n_pages": 190,
+  "est_tokens": 76000,
+  "expected_route": "openrouter",
+  "bundle_filenames": [
+    "source.pdf",
+    "hn_1.pdf",
+    "hn_2.pdf",
+    "hn_3.pdf",
+    "hn_4.pdf"
+  ],
+  "original_filenames": [
+    "2021-Apple-Catalog.pdf",
+    "web_6820ef1c6bfc9e7b.pdf",
+    "web_935425021e4da339.pdf",
+    "web_d939e3897d0c1a86.pdf",
+    "web_449e1fe2966b96f2.pdf"
+  ]
+}

q203/random_k2/question.json

@@ -0,0 +1,33 @@
+{
+  "cell_id": "q203_random_k2",
+  "doc_id": "2021-Apple-Catalog.pdf",
+  "question": "One40 can only be used for Apple Watch, is that true? Please directly answer 'yes' or 'no'.",
+  "answer": "Yes",
+  "answer_format": "Str",
+  "condition": "random",
+  "k": 2,
+  "source_position": "first",
+  "pdf_paths": [
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/pdfs/2021-Apple-Catalog.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_b62436644bb7bb60.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_2e63ac98af5701f4.pdf"
+  ],
+  "pdf_names": [
+    "2021-Apple-Catalog.pdf",
+    "web_b62436644bb7bb60.pdf",
+    "web_2e63ac98af5701f4.pdf"
+  ],
+  "n_pages": 150,
+  "est_tokens": 60000,
+  "expected_route": "openrouter",
+  "bundle_filenames": [
+    "source.pdf",
+    "random_1.pdf",
+    "random_2.pdf"
+  ],
+  "original_filenames": [
+    "2021-Apple-Catalog.pdf",
+    "web_b62436644bb7bb60.pdf",
+    "web_2e63ac98af5701f4.pdf"
+  ]
+}

q203/random_k4/question.json

@@ -0,0 +1,41 @@
+{
+  "cell_id": "q203_random_k4",
+  "doc_id": "2021-Apple-Catalog.pdf",
+  "question": "One40 can only be used for Apple Watch, is that true? Please directly answer 'yes' or 'no'.",
+  "answer": "Yes",
+  "answer_format": "Str",
+  "condition": "random",
+  "k": 4,
+  "source_position": "first",
+  "pdf_paths": [
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/pdfs/2021-Apple-Catalog.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_b62436644bb7bb60.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_2e63ac98af5701f4.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_ef8ec61b330e1c2a.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_1470bb8798a36da4.pdf"
+  ],
+  "pdf_names": [
+    "2021-Apple-Catalog.pdf",
+    "web_b62436644bb7bb60.pdf",
+    "web_2e63ac98af5701f4.pdf",
+    "web_ef8ec61b330e1c2a.pdf",
+    "web_1470bb8798a36da4.pdf"
+  ],
+  "n_pages": 276,
+  "est_tokens": 110400,
+  "expected_route": "openrouter",
+  "bundle_filenames": [
+    "source.pdf",
+    "random_1.pdf",
+    "random_2.pdf",
+    "random_3.pdf",
+    "random_4.pdf"
+  ],
+  "original_filenames": [
+    "2021-Apple-Catalog.pdf",
+    "web_b62436644bb7bb60.pdf",
+    "web_2e63ac98af5701f4.pdf",
+    "web_ef8ec61b330e1c2a.pdf",
+    "web_1470bb8798a36da4.pdf"
+  ]
+}

q204/control_k0/question.json

@@ -0,0 +1,25 @@
+{
+  "cell_id": "q204_control_k0",
+  "doc_id": "ACTIVISIONBLIZZARD_2019_10K.pdf",
+  "question": "what is the FY2019 fixed asset turnover ratio for Activision Blizzard?",
+  "answer": "24.26",
+  "answer_format": "Float",
+  "condition": "control",
+  "k": 0,
+  "source_position": "first",
+  "pdf_paths": [
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/pdfs/ACTIVISIONBLIZZARD_2019_10K.pdf"
+  ],
+  "pdf_names": [
+    "ACTIVISIONBLIZZARD_2019_10K.pdf"
+  ],
+  "n_pages": 198,
+  "est_tokens": 79200,
+  "expected_route": "openrouter",
+  "bundle_filenames": [
+    "source.pdf"
+  ],
+  "original_filenames": [
+    "ACTIVISIONBLIZZARD_2019_10K.pdf"
+  ]
+}

q204/hard_negative_k2/question.json

@@ -0,0 +1,33 @@
+{
+  "cell_id": "q204_hard_negative_k2",
+  "doc_id": "ACTIVISIONBLIZZARD_2019_10K.pdf",
+  "question": "what is the FY2019 fixed asset turnover ratio for Activision Blizzard?",
+  "answer": "24.26",
+  "answer_format": "Float",
+  "condition": "hard_negative",
+  "k": 2,
+  "source_position": "first",
+  "pdf_paths": [
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/pdfs/ACTIVISIONBLIZZARD_2019_10K.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_e3c6921f62a4a2ad.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_f7802e2b7d461d07.pdf"
+  ],
+  "pdf_names": [
+    "ACTIVISIONBLIZZARD_2019_10K.pdf",
+    "web_e3c6921f62a4a2ad.pdf",
+    "web_f7802e2b7d461d07.pdf"
+  ],
+  "n_pages": 354,
+  "est_tokens": 141600,
+  "expected_route": "openrouter",
+  "bundle_filenames": [
+    "source.pdf",
+    "hn_1.pdf",
+    "hn_2.pdf"
+  ],
+  "original_filenames": [
+    "ACTIVISIONBLIZZARD_2019_10K.pdf",
+    "web_e3c6921f62a4a2ad.pdf",
+    "web_f7802e2b7d461d07.pdf"
+  ]
+}

q204/hard_negative_k4/question.json

@@ -0,0 +1,41 @@
+{
+  "cell_id": "q204_hard_negative_k4",
+  "doc_id": "ACTIVISIONBLIZZARD_2019_10K.pdf",
+  "question": "what is the FY2019 fixed asset turnover ratio for Activision Blizzard?",
+  "answer": "24.26",
+  "answer_format": "Float",
+  "condition": "hard_negative",
+  "k": 4,
+  "source_position": "first",
+  "pdf_paths": [
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/pdfs/ACTIVISIONBLIZZARD_2019_10K.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_e3c6921f62a4a2ad.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_f7802e2b7d461d07.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_22ce47341478977c.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_dc179c2563f3e3ab.pdf"
+  ],
+  "pdf_names": [
+    "ACTIVISIONBLIZZARD_2019_10K.pdf",
+    "web_e3c6921f62a4a2ad.pdf",
+    "web_f7802e2b7d461d07.pdf",
+    "web_22ce47341478977c.pdf",
+    "web_dc179c2563f3e3ab.pdf"
+  ],
+  "n_pages": 595,
+  "est_tokens": 238000,
+  "expected_route": "openrouter",
+  "bundle_filenames": [
+    "source.pdf",
+    "hn_1.pdf",
+    "hn_2.pdf",
+    "hn_3.pdf",
+    "hn_4.pdf"
+  ],
+  "original_filenames": [
+    "ACTIVISIONBLIZZARD_2019_10K.pdf",
+    "web_e3c6921f62a4a2ad.pdf",
+    "web_f7802e2b7d461d07.pdf",
+    "web_22ce47341478977c.pdf",
+    "web_dc179c2563f3e3ab.pdf"
+  ]
+}

q204/random_k2/question.json

@@ -0,0 +1,33 @@
+{
+  "cell_id": "q204_random_k2",
+  "doc_id": "ACTIVISIONBLIZZARD_2019_10K.pdf",
+  "question": "what is the FY2019 fixed asset turnover ratio for Activision Blizzard?",
+  "answer": "24.26",
+  "answer_format": "Float",
+  "condition": "random",
+  "k": 2,
+  "source_position": "first",
+  "pdf_paths": [
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/pdfs/ACTIVISIONBLIZZARD_2019_10K.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/984e1ce91a6d9c30560e6c080bb62a819f8c2fe8.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_b44884342e7a9f39.pdf"
+  ],
+  "pdf_names": [
+    "ACTIVISIONBLIZZARD_2019_10K.pdf",
+    "984e1ce91a6d9c30560e6c080bb62a819f8c2fe8.pdf",
+    "web_b44884342e7a9f39.pdf"
+  ],
+  "n_pages": 312,
+  "est_tokens": 124800,
+  "expected_route": "openrouter",
+  "bundle_filenames": [
+    "source.pdf",
+    "random_1.pdf",
+    "random_2.pdf"
+  ],
+  "original_filenames": [
+    "ACTIVISIONBLIZZARD_2019_10K.pdf",
+    "984e1ce91a6d9c30560e6c080bb62a819f8c2fe8.pdf",
+    "web_b44884342e7a9f39.pdf"
+  ]
+}

q204/random_k4/question.json

@@ -0,0 +1,41 @@
+{
+  "cell_id": "q204_random_k4",
+  "doc_id": "ACTIVISIONBLIZZARD_2019_10K.pdf",
+  "question": "what is the FY2019 fixed asset turnover ratio for Activision Blizzard?",
+  "answer": "24.26",
+  "answer_format": "Float",
+  "condition": "random",
+  "k": 4,
+  "source_position": "first",
+  "pdf_paths": [
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/pdfs/ACTIVISIONBLIZZARD_2019_10K.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/984e1ce91a6d9c30560e6c080bb62a819f8c2fe8.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_b44884342e7a9f39.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_b873ee670af56beb.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_9942a44c5f6232cc.pdf"
+  ],
+  "pdf_names": [
+    "ACTIVISIONBLIZZARD_2019_10K.pdf",
+    "984e1ce91a6d9c30560e6c080bb62a819f8c2fe8.pdf",
+    "web_b44884342e7a9f39.pdf",
+    "web_b873ee670af56beb.pdf",
+    "web_9942a44c5f6232cc.pdf"
+  ],
+  "n_pages": 364,
+  "est_tokens": 145600,
+  "expected_route": "openrouter",
+  "bundle_filenames": [
+    "source.pdf",
+    "random_1.pdf",
+    "random_2.pdf",
+    "random_3.pdf",
+    "random_4.pdf"
+  ],
+  "original_filenames": [
+    "ACTIVISIONBLIZZARD_2019_10K.pdf",
+    "984e1ce91a6d9c30560e6c080bb62a819f8c2fe8.pdf",
+    "web_b44884342e7a9f39.pdf",
+    "web_b873ee670af56beb.pdf",
+    "web_9942a44c5f6232cc.pdf"
+  ]
+}

q235/control_k0/question.json

@@ -0,0 +1,25 @@
+{
+  "cell_id": "q235_control_k0",
+  "doc_id": "NETFLIX_2015_10K.pdf",
+  "question": "what profit  did Memberships contribute to in Domestic Streaming Segment in FY2015? Answer in thousands.",
+  "answer": "1375500.00",
+  "answer_format": "Float",
+  "condition": "control",
+  "k": 0,
+  "source_position": "first",
+  "pdf_paths": [
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/pdfs/NETFLIX_2015_10K.pdf"
+  ],
+  "pdf_names": [
+    "NETFLIX_2015_10K.pdf"
+  ],
+  "n_pages": 72,
+  "est_tokens": 28800,
+  "expected_route": "openrouter",
+  "bundle_filenames": [
+    "source.pdf"
+  ],
+  "original_filenames": [
+    "NETFLIX_2015_10K.pdf"
+  ]
+}

q235/hard_negative_k2/question.json

@@ -0,0 +1,33 @@
+{
+  "cell_id": "q235_hard_negative_k2",
+  "doc_id": "NETFLIX_2015_10K.pdf",
+  "question": "what profit  did Memberships contribute to in Domestic Streaming Segment in FY2015? Answer in thousands.",
+  "answer": "1375500.00",
+  "answer_format": "Float",
+  "condition": "hard_negative",
+  "k": 2,
+  "source_position": "first",
+  "pdf_paths": [
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/pdfs/NETFLIX_2015_10K.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_0cf212458514abec.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_252a096247d3374e.pdf"
+  ],
+  "pdf_names": [
+    "NETFLIX_2015_10K.pdf",
+    "web_0cf212458514abec.pdf",
+    "web_252a096247d3374e.pdf"
+  ],
+  "n_pages": 213,
+  "est_tokens": 85200,
+  "expected_route": "openrouter",
+  "bundle_filenames": [
+    "source.pdf",
+    "hn_1.pdf",
+    "hn_2.pdf"
+  ],
+  "original_filenames": [
+    "NETFLIX_2015_10K.pdf",
+    "web_0cf212458514abec.pdf",
+    "web_252a096247d3374e.pdf"
+  ]
+}

q235/hard_negative_k4/question.json

@@ -0,0 +1,41 @@
+{
+  "cell_id": "q235_hard_negative_k4",
+  "doc_id": "NETFLIX_2015_10K.pdf",
+  "question": "what profit  did Memberships contribute to in Domestic Streaming Segment in FY2015? Answer in thousands.",
+  "answer": "1375500.00",
+  "answer_format": "Float",
+  "condition": "hard_negative",
+  "k": 4,
+  "source_position": "first",
+  "pdf_paths": [
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/pdfs/NETFLIX_2015_10K.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_0cf212458514abec.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_252a096247d3374e.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_be2d7f897b96878c.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_e797ebb9acb0cdd4.pdf"
+  ],
+  "pdf_names": [
+    "NETFLIX_2015_10K.pdf",
+    "web_0cf212458514abec.pdf",
+    "web_252a096247d3374e.pdf",
+    "web_be2d7f897b96878c.pdf",
+    "web_e797ebb9acb0cdd4.pdf"
+  ],
+  "n_pages": 512,
+  "est_tokens": 204800,
+  "expected_route": "openrouter",
+  "bundle_filenames": [
+    "source.pdf",
+    "hn_1.pdf",
+    "hn_2.pdf",
+    "hn_3.pdf",
+    "hn_4.pdf"
+  ],
+  "original_filenames": [
+    "NETFLIX_2015_10K.pdf",
+    "web_0cf212458514abec.pdf",
+    "web_252a096247d3374e.pdf",
+    "web_be2d7f897b96878c.pdf",
+    "web_e797ebb9acb0cdd4.pdf"
+  ]
+}

q235/random_k2/question.json

@@ -0,0 +1,33 @@
+{
+  "cell_id": "q235_random_k2",
+  "doc_id": "NETFLIX_2015_10K.pdf",
+  "question": "what profit  did Memberships contribute to in Domestic Streaming Segment in FY2015? Answer in thousands.",
+  "answer": "1375500.00",
+  "answer_format": "Float",
+  "condition": "random",
+  "k": 2,
+  "source_position": "first",
+  "pdf_paths": [
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/pdfs/NETFLIX_2015_10K.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_c37ff5b2c1992920.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_26ccb3c73dadf54b.pdf"
+  ],
+  "pdf_names": [
+    "NETFLIX_2015_10K.pdf",
+    "web_c37ff5b2c1992920.pdf",
+    "web_26ccb3c73dadf54b.pdf"
+  ],
+  "n_pages": 166,
+  "est_tokens": 66400,
+  "expected_route": "openrouter",
+  "bundle_filenames": [
+    "source.pdf",
+    "random_1.pdf",
+    "random_2.pdf"
+  ],
+  "original_filenames": [
+    "NETFLIX_2015_10K.pdf",
+    "web_c37ff5b2c1992920.pdf",
+    "web_26ccb3c73dadf54b.pdf"
+  ]
+}

q235/random_k4/question.json

@@ -0,0 +1,41 @@
+{
+  "cell_id": "q235_random_k4",
+  "doc_id": "NETFLIX_2015_10K.pdf",
+  "question": "what profit  did Memberships contribute to in Domestic Streaming Segment in FY2015? Answer in thousands.",
+  "answer": "1375500.00",
+  "answer_format": "Float",
+  "condition": "random",
+  "k": 4,
+  "source_position": "first",
+  "pdf_paths": [
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/pdfs/NETFLIX_2015_10K.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_c37ff5b2c1992920.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_26ccb3c73dadf54b.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/5349694f4eecdd46092594856d170286c4d2fc94.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_54b7670fb9d9c193.pdf"
+  ],
+  "pdf_names": [
+    "NETFLIX_2015_10K.pdf",
+    "web_c37ff5b2c1992920.pdf",
+    "web_26ccb3c73dadf54b.pdf",
+    "5349694f4eecdd46092594856d170286c4d2fc94.pdf",
+    "web_54b7670fb9d9c193.pdf"
+  ],
+  "n_pages": 229,
+  "est_tokens": 91600,
+  "expected_route": "openrouter",
+  "bundle_filenames": [
+    "source.pdf",
+    "random_1.pdf",
+    "random_2.pdf",
+    "random_3.pdf",
+    "random_4.pdf"
+  ],
+  "original_filenames": [
+    "NETFLIX_2015_10K.pdf",
+    "web_c37ff5b2c1992920.pdf",
+    "web_26ccb3c73dadf54b.pdf",
+    "5349694f4eecdd46092594856d170286c4d2fc94.pdf",
+    "web_54b7670fb9d9c193.pdf"
+  ]
+}

q250/control_k0/question.json

@@ -0,0 +1,25 @@
+{
+  "cell_id": "q250_control_k0",
+  "doc_id": "AMAZON_2017_10K.pdf",
+  "question": "what is Amazon's FY2017 effective tax rate? round your answer to three decimal",
+  "answer": "0.202",
+  "answer_format": "Float",
+  "condition": "control",
+  "k": 0,
+  "source_position": "first",
+  "pdf_paths": [
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/pdfs/AMAZON_2017_10K.pdf"
+  ],
+  "pdf_names": [
+    "AMAZON_2017_10K.pdf"
+  ],
+  "n_pages": 85,
+  "est_tokens": 34000,
+  "expected_route": "openrouter",
+  "bundle_filenames": [
+    "source.pdf"
+  ],
+  "original_filenames": [
+    "AMAZON_2017_10K.pdf"
+  ]
+}

q250/hard_negative_k2/question.json

@@ -0,0 +1,33 @@
+{
+  "cell_id": "q250_hard_negative_k2",
+  "doc_id": "AMAZON_2017_10K.pdf",
+  "question": "what is Amazon's FY2017 effective tax rate? round your answer to three decimal",
+  "answer": "0.202",
+  "answer_format": "Float",
+  "condition": "hard_negative",
+  "k": 2,
+  "source_position": "first",
+  "pdf_paths": [
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/pdfs/AMAZON_2017_10K.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_a4c03fb8a963832d.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_707ad342af7f0e59.pdf"
+  ],
+  "pdf_names": [
+    "AMAZON_2017_10K.pdf",
+    "web_a4c03fb8a963832d.pdf",
+    "web_707ad342af7f0e59.pdf"
+  ],
+  "n_pages": 331,
+  "est_tokens": 132400,
+  "expected_route": "openrouter",
+  "bundle_filenames": [
+    "source.pdf",
+    "hn_1.pdf",
+    "hn_2.pdf"
+  ],
+  "original_filenames": [
+    "AMAZON_2017_10K.pdf",
+    "web_a4c03fb8a963832d.pdf",
+    "web_707ad342af7f0e59.pdf"
+  ]
+}

q250/hard_negative_k4/question.json

@@ -0,0 +1,41 @@
+{
+  "cell_id": "q250_hard_negative_k4",
+  "doc_id": "AMAZON_2017_10K.pdf",
+  "question": "what is Amazon's FY2017 effective tax rate? round your answer to three decimal",
+  "answer": "0.202",
+  "answer_format": "Float",
+  "condition": "hard_negative",
+  "k": 4,
+  "source_position": "first",
+  "pdf_paths": [
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/pdfs/AMAZON_2017_10K.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_a4c03fb8a963832d.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_707ad342af7f0e59.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_ca39b8721f6d1255.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_1d249d12039bf55b.pdf"
+  ],
+  "pdf_names": [
+    "AMAZON_2017_10K.pdf",
+    "web_a4c03fb8a963832d.pdf",
+    "web_707ad342af7f0e59.pdf",
+    "web_ca39b8721f6d1255.pdf",
+    "web_1d249d12039bf55b.pdf"
+  ],
+  "n_pages": 408,
+  "est_tokens": 163200,
+  "expected_route": "openrouter",
+  "bundle_filenames": [
+    "source.pdf",
+    "hn_1.pdf",
+    "hn_2.pdf",
+    "hn_3.pdf",
+    "hn_4.pdf"
+  ],
+  "original_filenames": [
+    "AMAZON_2017_10K.pdf",
+    "web_a4c03fb8a963832d.pdf",
+    "web_707ad342af7f0e59.pdf",
+    "web_ca39b8721f6d1255.pdf",
+    "web_1d249d12039bf55b.pdf"
+  ]
+}

q250/random_k2/question.json

@@ -0,0 +1,33 @@
+{
+  "cell_id": "q250_random_k2",
+  "doc_id": "AMAZON_2017_10K.pdf",
+  "question": "what is Amazon's FY2017 effective tax rate? round your answer to three decimal",
+  "answer": "0.202",
+  "answer_format": "Float",
+  "condition": "random",
+  "k": 2,
+  "source_position": "first",
+  "pdf_paths": [
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/pdfs/AMAZON_2017_10K.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_6bc8f0013b78aca4.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_f253c9f7d356a5ba.pdf"
+  ],
+  "pdf_names": [
+    "AMAZON_2017_10K.pdf",
+    "web_6bc8f0013b78aca4.pdf",
+    "web_f253c9f7d356a5ba.pdf"
+  ],
+  "n_pages": 165,
+  "est_tokens": 66000,
+  "expected_route": "openrouter",
+  "bundle_filenames": [
+    "source.pdf",
+    "random_1.pdf",
+    "random_2.pdf"
+  ],
+  "original_filenames": [
+    "AMAZON_2017_10K.pdf",
+    "web_6bc8f0013b78aca4.pdf",
+    "web_f253c9f7d356a5ba.pdf"
+  ]
+}

q250/random_k4/question.json

@@ -0,0 +1,41 @@
+{
+  "cell_id": "q250_random_k4",
+  "doc_id": "AMAZON_2017_10K.pdf",
+  "question": "what is Amazon's FY2017 effective tax rate? round your answer to three decimal",
+  "answer": "0.202",
+  "answer_format": "Float",
+  "condition": "random",
+  "k": 4,
+  "source_position": "first",
+  "pdf_paths": [
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/pdfs/AMAZON_2017_10K.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_6bc8f0013b78aca4.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_f253c9f7d356a5ba.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_28449e05c7149be2.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_826aff3b5899abff.pdf"
+  ],
+  "pdf_names": [
+    "AMAZON_2017_10K.pdf",
+    "web_6bc8f0013b78aca4.pdf",
+    "web_f253c9f7d356a5ba.pdf",
+    "web_28449e05c7149be2.pdf",
+    "web_826aff3b5899abff.pdf"
+  ],
+  "n_pages": 200,
+  "est_tokens": 80000,
+  "expected_route": "openrouter",
+  "bundle_filenames": [
+    "source.pdf",
+    "random_1.pdf",
+    "random_2.pdf",
+    "random_3.pdf",
+    "random_4.pdf"
+  ],
+  "original_filenames": [
+    "AMAZON_2017_10K.pdf",
+    "web_6bc8f0013b78aca4.pdf",
+    "web_f253c9f7d356a5ba.pdf",
+    "web_28449e05c7149be2.pdf",
+    "web_826aff3b5899abff.pdf"
+  ]
+}

q268/control_k0/question.json

@@ -0,0 +1,25 @@
+{
+  "cell_id": "q268_control_k0",
+  "doc_id": "BESTBUY_2023_10K.pdf",
+  "question": "what is After-tax Return on Average Equity  for the fiscal year ending January 28, 2023? round your answer to three decimal places",
+  "answer": "0.488",
+  "answer_format": "Float",
+  "condition": "control",
+  "k": 0,
+  "source_position": "first",
+  "pdf_paths": [
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/pdfs/BESTBUY_2023_10K.pdf"
+  ],
+  "pdf_names": [
+    "BESTBUY_2023_10K.pdf"
+  ],
+  "n_pages": 75,
+  "est_tokens": 30000,
+  "expected_route": "openrouter",
+  "bundle_filenames": [
+    "source.pdf"
+  ],
+  "original_filenames": [
+    "BESTBUY_2023_10K.pdf"
+  ]
+}

q268/hard_negative_k2/question.json

@@ -0,0 +1,33 @@
+{
+  "cell_id": "q268_hard_negative_k2",
+  "doc_id": "BESTBUY_2023_10K.pdf",
+  "question": "what is After-tax Return on Average Equity  for the fiscal year ending January 28, 2023? round your answer to three decimal places",
+  "answer": "0.488",
+  "answer_format": "Float",
+  "condition": "hard_negative",
+  "k": 2,
+  "source_position": "first",
+  "pdf_paths": [
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/pdfs/BESTBUY_2023_10K.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_cd0ef8929be1ba45.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_7d19f6428d35b598.pdf"
+  ],
+  "pdf_names": [
+    "BESTBUY_2023_10K.pdf",
+    "web_cd0ef8929be1ba45.pdf",
+    "web_7d19f6428d35b598.pdf"
+  ],
+  "n_pages": 255,
+  "est_tokens": 102000,
+  "expected_route": "openrouter",
+  "bundle_filenames": [
+    "source.pdf",
+    "hn_1.pdf",
+    "hn_2.pdf"
+  ],
+  "original_filenames": [
+    "BESTBUY_2023_10K.pdf",
+    "web_cd0ef8929be1ba45.pdf",
+    "web_7d19f6428d35b598.pdf"
+  ]
+}

q268/hard_negative_k4/question.json

@@ -0,0 +1,41 @@
+{
+  "cell_id": "q268_hard_negative_k4",
+  "doc_id": "BESTBUY_2023_10K.pdf",
+  "question": "what is After-tax Return on Average Equity  for the fiscal year ending January 28, 2023? round your answer to three decimal places",
+  "answer": "0.488",
+  "answer_format": "Float",
+  "condition": "hard_negative",
+  "k": 4,
+  "source_position": "first",
+  "pdf_paths": [
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/pdfs/BESTBUY_2023_10K.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_cd0ef8929be1ba45.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_7d19f6428d35b598.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_d799da399f7e56a2.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_4eb4afea574790a6.pdf"
+  ],
+  "pdf_names": [
+    "BESTBUY_2023_10K.pdf",
+    "web_cd0ef8929be1ba45.pdf",
+    "web_7d19f6428d35b598.pdf",
+    "web_d799da399f7e56a2.pdf",
+    "web_4eb4afea574790a6.pdf"
+  ],
+  "n_pages": 427,
+  "est_tokens": 170800,
+  "expected_route": "openrouter",
+  "bundle_filenames": [
+    "source.pdf",
+    "hn_1.pdf",
+    "hn_2.pdf",
+    "hn_3.pdf",
+    "hn_4.pdf"
+  ],
+  "original_filenames": [
+    "BESTBUY_2023_10K.pdf",
+    "web_cd0ef8929be1ba45.pdf",
+    "web_7d19f6428d35b598.pdf",
+    "web_d799da399f7e56a2.pdf",
+    "web_4eb4afea574790a6.pdf"
+  ]
+}

q268/random_k2/question.json

@@ -0,0 +1,33 @@
+{
+  "cell_id": "q268_random_k2",
+  "doc_id": "BESTBUY_2023_10K.pdf",
+  "question": "what is After-tax Return on Average Equity  for the fiscal year ending January 28, 2023? round your answer to three decimal places",
+  "answer": "0.488",
+  "answer_format": "Float",
+  "condition": "random",
+  "k": 2,
+  "source_position": "first",
+  "pdf_paths": [
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/pdfs/BESTBUY_2023_10K.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_c00a826ae0d1be47.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_c3b929d619aa472c.pdf"
+  ],
+  "pdf_names": [
+    "BESTBUY_2023_10K.pdf",
+    "web_c00a826ae0d1be47.pdf",
+    "web_c3b929d619aa472c.pdf"
+  ],
+  "n_pages": 181,
+  "est_tokens": 72400,
+  "expected_route": "openrouter",
+  "bundle_filenames": [
+    "source.pdf",
+    "random_1.pdf",
+    "random_2.pdf"
+  ],
+  "original_filenames": [
+    "BESTBUY_2023_10K.pdf",
+    "web_c00a826ae0d1be47.pdf",
+    "web_c3b929d619aa472c.pdf"
+  ]
+}

q268/random_k4/question.json

@@ -0,0 +1,41 @@
+{
+  "cell_id": "q268_random_k4",
+  "doc_id": "BESTBUY_2023_10K.pdf",
+  "question": "what is After-tax Return on Average Equity  for the fiscal year ending January 28, 2023? round your answer to three decimal places",
+  "answer": "0.488",
+  "answer_format": "Float",
+  "condition": "random",
+  "k": 4,
+  "source_position": "first",
+  "pdf_paths": [
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/pdfs/BESTBUY_2023_10K.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_c00a826ae0d1be47.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_c3b929d619aa472c.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_999c7640a084a7b3.pdf",
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/hard_negative_pdfs/web_a72eb2d4f1801801.pdf"
+  ],
+  "pdf_names": [
+    "BESTBUY_2023_10K.pdf",
+    "web_c00a826ae0d1be47.pdf",
+    "web_c3b929d619aa472c.pdf",
+    "web_999c7640a084a7b3.pdf",
+    "web_a72eb2d4f1801801.pdf"
+  ],
+  "n_pages": 314,
+  "est_tokens": 125600,
+  "expected_route": "openrouter",
+  "bundle_filenames": [
+    "source.pdf",
+    "random_1.pdf",
+    "random_2.pdf",
+    "random_3.pdf",
+    "random_4.pdf"
+  ],
+  "original_filenames": [
+    "BESTBUY_2023_10K.pdf",
+    "web_c00a826ae0d1be47.pdf",
+    "web_c3b929d619aa472c.pdf",
+    "web_999c7640a084a7b3.pdf",
+    "web_a72eb2d4f1801801.pdf"
+  ]
+}

q29/control_k0/question.json

@@ -0,0 +1,25 @@
+{
+  "cell_id": "q29_control_k0",
+  "doc_id": "379f44022bb27aa53efd5d322c7b57bf.pdf",
+  "question": "List the primary questions asked about the services in this report.",
+  "answer": "['Is the servife safe?', 'Is the service effective', 'Is the serve caring?', 'Is the service responsive?', 'Is the service well-led?']",
+  "answer_format": "List",
+  "condition": "control",
+  "k": 0,
+  "source_position": "first",
+  "pdf_paths": [
+    "/Users/jasonluo08/Desktop/AI Brain/algoverse_research/pdfs/379f44022bb27aa53efd5d322c7b57bf.pdf"
+  ],
+  "pdf_names": [
+    "379f44022bb27aa53efd5d322c7b57bf.pdf"
+  ],
+  "n_pages": 17,
+  "est_tokens": 6800,
+  "expected_route": "azure",
+  "bundle_filenames": [
+    "source.pdf"
+  ],
+  "original_filenames": [
+    "379f44022bb27aa53efd5d322c7b57bf.pdf"
+  ]
+}