Answer quality: 100 questions, scored in public

One hundred questions a patient, carer, clinician, or analyst might ask, each with a grounded expected answer and a rubric of must-mention points. The same rubric scores OnCo, a search engine, or an AI assistant. Scores are published here in public, every run.

Questions

100

69 factual · 7 procedural · 24 reasoning

OnCo latest score

84%

2026-09-08 · retrieval recall 45%

Systems scored

3 runs on file

Leaderboard

System	Date	Mean	Factual	Procedural	Reasoning	Method
OnCo (search-retrieval proxy, top 8 entities)	2026-09-08	84%	86%	62%	85%	For each question: MiniSearch retrieval over the site index; rubric coverage of tldr+summary of the top entities. 1 point per rubric item met, normalised to 0-1.
OnCo Ask (intents, entity resolution, templated cited answers; extractive fallback)	2026-09-10	83%	87%	64%	76%	For each question: intent from wording rules; records named in the question resolved by longest match over names and aliases; a template per intent fills sentences from those records' structured fields (TL;DR first), each sentence cited; curated benchmark and patient-question pairs supply records when the question matches closely; sentence retrieval (MiniSearch plus concept index, fused) is the fallback and adds detail. The answer text shown on /ask/ is scored for rubric coverage. No language model.
OnCo Ask (extractive answer, lexical + concept retrieval, top 6 records)	2026-09-09	71%	76%	44%	66%	For each question: MiniSearch and concept-index retrieval fused by reciprocal rank, top 6 records; the extractive answer shown on /ask/ (best sentences, two per record, one citation each) is scored for rubric coverage. No language model.

OnCo's own run is a deliberately hard retrieval proxy: it scores only the TL;DR and summary text of the top eight search hits, with no reasoning layer. A human reading the linked pages would score higher; that gap is part of what the benchmark measures.

Score another system

1. Download the questions: benchmark.json (also in the repo at src/data/benchmark.ts).

2. Ask each question to the system under test, verbatim, and save the answers as { "<question id>": "answer text", … }.

3. Run npm run bench:score -- answers.json "System name" and open a pull request with the generated file in public/eval/. The rubric is public; the only judgement is substring matching, so scores are reproducible and disputable.

The questions

100 questions

	Question			Grounded in
tnbc-01	Which TROP2 ADCs are approved for first-line metastatic triple-negative breast cancer? Expected: Sacituzumab govitecan (Trodelvy), as monotherapy for PD-1-ineligible patients (ASCENT-03) and with pembrolizumab for PD-L1 CPS ≥10 disease (ASCENT-04), and datopotamab deruxtecan (Datroway) for PD-1/PD-L1-ineligible patients (TROPION-Breast02), both approved in 2026.	Factual	1	Sacituzumab govitecan, Datopotamab deruxtecan, ASCENT-03, ASCENT-04 / KEYNOTE-D19, TROPION-Breast02, Metastatic triple-negative breast cancer	4/4	All met
tnbc-02	What does 'triple-negative' mean in breast cancer? Expected: The tumour lacks the three receptors other breast cancers can be treated through: oestrogen receptor, progesterone receptor, and HER2 (IHC 0-1+ or 2+/ISH-negative).	Factual	1	Triple-negative breast cancer (TNBC), HER2, Estrogen receptor (ERα)	4/4	All met
tnbc-03	What is the standard treatment for stage II-III triple-negative breast cancer today? Expected: Neoadjuvant pembrolizumab with carboplatin/paclitaxel then anthracycline chemotherapy, surgery, and adjuvant pembrolizumab (KEYNOTE-522); adjuvant olaparib for germline BRCA carriers with residual disease (OlympiA); capecitabine for residual disease without BRCA.	Factual	2	Triple-negative breast cancer (TNBC), Early triple-negative breast cancer, KEYNOTE-522, Pembrolizumab, Olaparib, OlympiA	5/5	All met
tnbc-04	Did KEYNOTE-522 improve overall survival? Expected: Yes. Seven-year overall survival was 85.1% vs 77.2% (ASCO 2026 update); five-year EFS 81.2% vs 72.2%; pCR 64.8% vs 51.2%.	Factual	2	KEYNOTE-522	4/4	All met
tnbc-05	What is pathologic complete response and why does it matter in TNBC? Expected: pCR means no invasive cancer is left in the breast and lymph nodes when the surgeon removes the tissue after pre-surgery treatment. In TNBC patients with pCR have around 90% five-year event-free survival, and trials now test giving less treatment after pCR.	Factual	1	Pathologic complete response (pCR), OptimICE-pCR (A012103), Triple-negative breast cancer (TNBC)	4/4	All met
tnbc-06	Which trial showed the first overall survival benefit for a first-line ADC in triple-negative breast cancer? Expected: TROPION-Breast02: datopotamab deruxtecan vs chemotherapy in first-line PD-1/PD-L1-ineligible metastatic TNBC, OS 23.7 vs 18.7 months.	Factual	2	TROPION-Breast02, Datopotamab deruxtecan	4/4	All met
tnbc-07	What is the first bispecific ADC to succeed in a phase 3 trial, and in which cancer? Expected: Izalontamab brengitecan (iza-bren, BL-B01D1), an EGFR×HER3 bispecific ADC from SystImmune/BMS, met PFS and OS in previously treated TNBC (BL-B01D1-307, February 2026) and also in oesophageal squamous cell carcinoma.	Factual	2	Izalontamab brengitecan, BL-B01D1-307, Bispecific ADC	5/5	All met
tnbc-08	Why does trastuzumab deruxtecan work in 'HER2-low' breast cancers that older HER2 drugs ignored? Expected: Its cleavable linker releases a membrane-permeable topoisomerase-I payload (DXd) at high DAR, so a small amount of HER2 is enough to deliver drug and the payload diffuses to kill neighbouring cells (bystander effect); HER2-low is a delivery address, not a driver. DESTINY-Breast04 and -06 proved it.	Reasoning	2	Trastuzumab deruxtecan, HER2-low and HER2-ultralow, Bystander effect (ADC), DESTINY-Breast04	4/4	All met
tnbc-09	Should a patient get a second ADC with the same kind of payload straight after the first one fails? Expected: Caution: retrospective series (SATEEN, BRE-354) show shorter PFS for a second TOP1-payload ADC given back-to-back; resistance is at the payload level (SLFN11 loss, TOP1 mutations, ABCG2 efflux). Some data suggest interposing chemotherapy or switching payload class; prospective trials (TRADE-DXd) are pending.	Reasoning	3	Caution: TOP1 ADC immediately after TOP1 ADC, ADC sequencing, Payload-class switching as the rule for ADC sequencing	4/4	All met
tnbc-10	What questions should someone newly diagnosed with triple-negative breast cancer ask before surgery? Expected: Whether chemo-immunotherapy before surgery (KEYNOTE-522) is planned and why; germline BRCA testing; PD-L1 and HER2-low status; TIL score; clinical trial options; fertility preservation; breast-conserving versus mastectomy and sentinel node approach; what response (pCR/RCB) will mean for treatment afterwards.	Procedural	2	Triple-negative breast cancer (TNBC), Early triple-negative breast cancer, KEYNOTE-522, Germline (hereditary) testing, Tumour-infiltrating lymphocytes (TILs), Pathologic complete response (pCR), Residual cancer burden (RCB)	4/5	Partly met
tnbc-11	Does the TROP2 level measured on a biopsy tell you whether Trodelvy will work? Expected: Not reliably. ASCENT showed benefit regardless of TROP2 IHC, so no companion diagnostic is required; expression is heterogeneous and archival tissue may not reflect current status. TROP2 PET tracers are being developed to map expression across the body and over time.	Reasoning	3	TROP2, ASCENT, TROP2 PET, TROP2 PET to choose and sequence TROP2 ADCs	3/4	Partly met
breast-12	What was the first PROTAC ever approved and for what? Expected: Vepdegestrant (Veppanu, Arvinas/Pfizer), an oral oestrogen receptor degrader, approved in 2026 for ESR1-mutated ER+/HER2- advanced breast cancer (VERITAC-2).	Factual	2	Vepdegestrant, PROTACs & molecular glues (targeted protein degradation), VERITAC-2	4/4	All met
breast-13	Which CDK4/6 inhibitors are approved after surgery for early hormone-positive breast cancer, and on what trials? Expected: Abemaciclib (monarchE, high-risk node-positive) and ribociclib (NATALEE, stage II-III including node-negative).	Factual	2	Abemaciclib, Ribociclib, monarchE, NATALEE, High-risk early HR-positive breast cancer	4/4	All met
breast-14	What can a 21-gene test tell a woman with early hormone-positive breast cancer? Expected: Oncotype DX gives a recurrence score; TAILORx showed women with a score of 25 or below (over 50) get no benefit from chemotherapy and can safely skip it; RxPONDER extended this to 1-3 positive nodes in postmenopausal women.	Factual	1	Oncotype DX, HR-positive / HER2-negative breast cancer, High-risk early HR-positive breast cancer	4/4	All met
breast-15	In what setting did T-DXd move into early-stage HER2-positive breast cancer in 2026? Expected: Neoadjuvant (DESTINY-Breast11, T-DXd followed by THP, pCR 67.3% vs 56.3%) and post-neoadjuvant residual disease (DESTINY-Breast05, beating T-DM1), both approved Q2 2026.	Factual	2	Trastuzumab deruxtecan, DESTINY-Breast11, HER2-positive breast cancer, Early HER2-positive breast cancer	4/4	All met
lung-16	Which biomarkers must be tested at diagnosis of advanced non-small-cell lung cancer? Expected: EGFR, ALK, ROS1, BRAF V600E, MET exon 14/amplification, RET, NTRK, KRAS G12C, HER2 mutations, and PD-L1 TPS.	Factual	2	Non-small-cell lung cancer, EGFR, ALK, KRAS, PD-L1	5/5	All met
lung-17	What is the standard adjuvant treatment after resection of EGFR-mutant lung cancer and what did it do to survival? Expected: Three years of osimertinib (ADAURA): DFS HR 0.20 in stage II-IIIA and OS HR 0.49, five-year OS 88% vs 78%.	Factual	2	Osimertinib, ADAURA	1/4	Partly met
lung-18	Which regimen beat osimertinib in first-line EGFR-mutant lung cancer? Expected: Amivantamab plus lazertinib (MARIPOSA): PFS 23.7 vs 16.6 months and an OS benefit of more than a year.	Factual	2	Amivantamab, MARIPOSA	4/4	All met
lung-19	What is the first T-cell engager to improve survival in a common solid tumour? Expected: Tarlatamab (Imdelltra), DLL3×CD3, in second-line small-cell lung cancer: OS 13.6 vs 8.3 months in DeLLphi-304.	Factual	2	Tarlatamab, DeLLphi-304, Small-cell lung cancer	4/4	All met
lung-20	Why did the first KRAS inhibitors need an EGFR antibody added in colorectal cancer but not in lung cancer? Expected: Colorectal epithelium has strong EGFR-driven adaptive feedback that reactivates MAPK within hours of KRAS G12C inhibition, so monotherapy response is ~10%; adding cetuximab or panitumumab (CodeBreaK 300, KRYSTAL-1) raises it to 30-45%.	Reasoning	3	KRAS G12C inhibitor + anti-EGFR antibody (colorectal), CodeBreaK 300, Sotorasib, Adagrasib	3/4	Partly met
lung-21	Which drug beat pembrolizumab head to head on progression-free survival in lung cancer? Expected: Ivonescimab, a PD-1×VEGF bispecific (Akeso/Summit): PFS 11.1 vs 5.8 months in PD-L1-positive first-line NSCLC (HARMONi-2, China).	Factual	2	Ivonescimab, VEGF / VEGFR, PD-1	4/4	All met
lung-22	What is the longest progression-free survival ever reported for a targeted pill in metastatic lung cancer? Expected: Lorlatinib in ALK-positive NSCLC (CROWN): five-year PFS about 60% versus 8% for crizotinib.	Factual	2	Lorlatinib, ALK	4/4	All met
prostate-23	What is theranostics and what is the best-known example? Expected: Using the same targeting molecule for a diagnostic scan and a treatment: PSMA PET shows where prostate cancer is, and 177Lu-PSMA-617 (Pluvicto) delivers radiation to the same target.	Factual	1	Theranostics, PSMA PET, Lutetium-177 vipivotide tetraxetan, PSMA	4/4	All met
prostate-24	What did the VISION trial show? Expected: 177Lu-PSMA-617 plus standard care improved overall survival (15.3 vs 11.3 months, HR 0.62) in PSMA-positive metastatic castration-resistant prostate cancer after ARPI and taxane.	Factual	2	VISION, Lutetium-177 vipivotide tetraxetan	3/4	Partly met
prostate-25	Why might an alpha-emitting PSMA drug work after lutetium PSMA has stopped working? Expected: Alpha particles (actinium-225) deposit far more energy over 50-100 µm, causing clustered double-strand breaks independent of oxygen and cell cycle, so beta-resistant, hypoxic, or small-volume disease can still be killed; retrospective series show PSA responses in about half after 177Lu failure. Supply of Ac-225 and salivary toxicity are the limits.	Reasoning	3	Beta radioligand → alpha radioligand, Targeted alpha therapy, Actinium-225 PSMA agents, Alpha vs beta emitters	4/4	All met
prostate-26	What is the first FDA-cleared AI pathology tool that predicts treatment benefit, and in which cancer? Expected: ArteraAI Prostate (de novo authorisation, August 2025) predicts prognosis and benefit from short-term androgen deprivation with radiotherapy in localised prostate cancer.	Factual	2	ArteraAI Prostate, Digital pathology & AI	4/4	All met
bladder-27	What regimen replaced platinum chemotherapy as first-line treatment for advanced bladder cancer, and how big was the effect? Expected: Enfortumab vedotin plus pembrolizumab (EV-302): OS 31.5 vs 16.1 months, HR 0.47.	Factual	2	Enfortumab vedotin, EV-302 / KEYNOTE-A39, Bladder & urothelial cancer	4/4	All met
bladder-28	What was the first cancer drug approved on the basis of a blood test for leftover disease? Expected: Atezolizumab for ctDNA-positive muscle-invasive bladder cancer after cystectomy (IMvigor011, using Signatera), approved Q2 2026; DFS HR 0.64, OS HR 0.59.	Factual	2	Atezolizumab, IMvigor011, Signatera, MRD / molecular residual disease testing	4/4	All met
gi-29	Which cancer gene was called undruggable for forty years and what changed? Expected: KRAS. In 2013 a covalent pocket was found in the G12C mutant; sotorasib (2021) and adagrasib followed, and pan-RAS(ON) inhibitors such as daraxonrasib are now in phase 3 for pancreatic cancer.	Factual	2	KRAS, KRAS & RAS inhibitors, Sotorasib, Daraxonrasib, KRAS roadmap: undruggable → G12C → pan-RAS	4/4	All met
gi-30	What is daraxonrasib and where does it stand? Expected: Revolution Medicines' pan-RAS(ON) tri-complex inhibitor; phase 3 RASolute 302 in second-line pancreatic cancer (reported OS benefit, HR 0.40 per the pancreatic cancer deep dive), with first-line and NSCLC phase 3 trials ongoing.	Factual	2	Daraxonrasib, Revolution Medicines, Pancreatic ductal adenocarcinoma	4/4	All met
gi-31	What did the dostarlimab rectal cancer study show? Expected: In mismatch-repair-deficient locally advanced rectal cancer, dostarlimab alone produced a complete clinical response in 100% of patients, sustained in more than 40 patients by 2025, allowing surgery and radiation to be avoided.	Factual	2	Dostarlimab, Microsatellite instability (MSI-H) / mismatch repair deficiency (dMMR), Colorectal cancer	4/4	All met
gi-32	Can a blood test decide who needs chemotherapy after colon cancer surgery? Expected: Yes in stage II: the DYNAMIC trial used ctDNA to guide adjuvant chemotherapy, halving its use (15% vs 28%) with non-inferior recurrence-free survival (93.5% vs 92.4% at two years).	Reasoning	2	DYNAMIC, MRD / molecular residual disease testing, Circulating tumour DNA (ctDNA)	4/4	All met
gi-33	What was the first approval in nearly thirty years specific to locally advanced pancreatic cancer? Expected: Optune Pax (tumour treating fields) with gemcitabine/nab-paclitaxel, approved Q1 2026 on PANOVA-3.	Factual	2	Optune / Optune Pax (TTFields), Tumour treating fields (TTFields), Pancreatic ductal adenocarcinoma	4/4	All met
gi-34	Which new target has an approved antibody, an approved CAR-T in China, and ADCs in phase 3 for gastric cancer? Expected: Claudin 18.2: zolbetuximab (Vyloy), satricabtagene autoleucel (satri-cel, CARsgen, China), and ADCs such as CMG901/AZD0901.	Factual	2	Claudin 18.2, Satricabtagene autoleucel, Sonesitatug vedotin, Gastric & gastro-oesophageal junction cancer	4/4	All met
gi-35	Why is pancreatic cancer so hard to treat with immunotherapy? Expected: It is immunologically cold: dense desmoplastic stroma blocks drug and T-cell entry, low mutational burden gives few neoantigens, and an immunosuppressive myeloid microenvironment. Vaccines (autogene cevumeran) and RAS inhibitors that may increase antigen presentation are the main hopes.	Reasoning	3	Pancreatic ductal adenocarcinoma, Hot vs cold tumours, Autogene cevumeran, FAP	1/4	Partly met
io-36	What fraction of advanced melanoma patients on nivolumab plus ipilimumab are alive at ten years? Expected: About 43% overall survival (CheckMate 067), with melanoma-specific survival around 52%.	Factual	2	CheckMate 067, Nivolumab, Ipilimumab	4/4	All met
io-37	What was the first personalised cancer vaccine to win a phase 3 trial? Expected: Intismeran autogene (V940/mRNA-4157, Moderna/Merck) with pembrolizumab in resected stage IIB-IV melanoma: INTerpath-001 met recurrence-free and distant-metastasis-free survival, announced 19 August 2026.	Factual	2	Intismeran autogene, INTerpath-001 (V940-001), Personalised neoantigen (mRNA) vaccines	4/4	All met
io-38	How is a personalised mRNA cancer vaccine made? Expected: The tumour and normal DNA are sequenced, software predicts up to 34 neoantigens unique to the tumour, a patient-specific mRNA encoding them is manufactured in lipid nanoparticles over about six weeks, and it is given with a PD-1 blocker.	Factual	2	Personalised neoantigen (mRNA) vaccines, Neoantigen, Intismeran autogene	4/5	Partly met
io-39	What is the first approved TIL therapy and how well does it work? Expected: Lifileucel (Amtagvi, Iovance), approved 2024 for anti-PD-1-refractory melanoma: response rate about 31%, with roughly a third of responders still responding at five years.	Factual	2	Lifileucel, TIL therapy	4/4	All met
io-40	What are the main side effects of checkpoint inhibitors and how are they handled? Expected: Immune-related adverse events such as colitis, thyroid problems, rash, hepatitis, and pneumonitis, most common with CTLA-4 plus PD-1 combinations; treated with steroids and sometimes other immunosuppressants, some endocrine effects are permanent.	Procedural	2	Immune-related adverse events (irAEs), Immune checkpoint inhibitors	3/4	Partly met
io-41	Why did TIGIT-blocking antibodies fail despite promising early data? Expected: Phase 3 trials (SKYSCRAPER-01, -02 and others) did not add benefit to PD-(L)1 blockade in NSCLC and SCLC; phase 2 signals did not reproduce, and the roles of Fc-effector function and patient selection remain debated.	Reasoning	3	TIGIT, Caution: TIGIT + PD-(L)1 blockade	4/4	All met
io-42	What does 'hot' versus 'cold' tumour mean? Expected: Hot tumours are full of immune cells and tend to respond to immunotherapy; cold tumours have kept the immune system out (immune-excluded or immune-desert), such as pancreatic, prostate, glioblastoma, and most HR+ breast cancers.	Factual	1	Hot vs cold tumours	4/4	All met
heme-43	What is CAR-T and which cancers is it approved for? Expected: A patient's T cells are engineered with a synthetic receptor, multiplied, and returned. Approved CD19 products treat B-cell lymphomas and leukaemias; BCMA products treat multiple myeloma; the first solid-tumour approval (Claudin 18.2, gastric) is in China.	Factual	1	CAR-T cell therapy, CD19, BCMA	4/4	All met
heme-44	What were the first checkpoint inhibitor and the first ADC ever approved? Expected: Ipilimumab (2011) was the first checkpoint inhibitor; gemtuzumab ozogamicin (Mylotarg, 2000) was the first ADC, withdrawn in 2010 and re-approved in 2017.	Factual	2	Ipilimumab, Gemtuzumab ozogamicin, ADC roadmap: from Mylotarg to bispecific and dual-payload ADCs	2/4	Partly met
heme-45	What is the first menin inhibitor and who is it for? Expected: Revumenib (Revuforj, Syndax), approved 2024 for relapsed KMT2A-rearranged acute leukaemia and 2025 for NPM1-mutant AML.	Factual	2	Revumenib, Menin	4/4	All met
heme-46	How did CAR-T change second-line treatment of large B-cell lymphoma? Expected: ZUMA-7 showed axicabtagene ciloleucel beats standard chemotherapy plus transplant for early relapse; CAR-T is now second-line standard, curing around 40% of relapsed patients.	Factual	2	Axicabtagene ciloleucel, Diffuse large B-cell lymphoma	4/4	All met
heme-47	Why does venetoclax work in leukaemia? Expected: It blocks BCL-2, the protein that stops leukaemia cells from self-destructing, so the built-in death programme (apoptosis) can run; used in CLL (fixed duration with obinutuzumab) and AML (with azacitidine).	Reasoning	2	Venetoclax, BCL-2, Intrinsic apoptosis (BCL-2 family)	2/4	Partly met
img-48	What is the difference between a CT scan and a PET scan? Expected: CT is a fast 3D X-ray showing size and shape; PET shows biology by tracking where a radioactive tracer accumulates (for example glucose uptake with FDG or a specific protein such as PSMA). PET/CT combines both.	Factual	1	CT (computed tomography), PET (positron emission tomography), PET/CT, FDG PET	3/4	Partly met
img-49	What does a FAPI PET scan see that an FDG scan often misses? Expected: The activated fibroblasts (FAP) in tumour stroma, present in more than 90% of epithelial cancers, giving high contrast in pancreatic, gastric, HCC, and peritoneal disease where FDG is weak; it is not yet approved.	Factual	2	FAPI PET, FAP	4/4	All met
img-50	What is the Galleri test and where does it stand with the FDA? Expected: GRAIL's methylation-based multi-cancer early detection blood test for more than 50 cancers; PMA submitted January 2026 on PATHFINDER 2 and NHS-Galleri data, FDA advisory committee 23 September 2026. NHS-Galleri missed its primary stage III-IV reduction endpoint (stage IV fell ~14%).	Factual	2	Galleri, Multi-cancer early detection (MCED), NHS-Galleri, PATHFINDER 2	4/4	All met
img-51	What is positive predictive value and why does it matter for a cancer screening blood test? Expected: The chance that a positive result is truly cancer. With a low-prevalence disease even a highly specific test yields many false positives; Galleri's PPV in PATHFINDER 2 was around 40-60%, so roughly half of positives lead to a diagnostic workup that finds no cancer.	Reasoning	2	Positive predictive value (PPV), Multi-cancer early detection (MCED), Galleri	4/4	All met
img-52	What is a liquid biopsy used for in cancer care today? Expected: A blood test reading tumour DNA fragments: to genotype a tumour when tissue is scarce, to track resistance mutations (EGFR T790M, ESR1), to detect minimal residual disease after surgery, and, experimentally, to screen for cancer.	Factual	1	Liquid biopsy (ctDNA), Circulating tumour DNA (ctDNA), MRD / molecular residual disease testing	2/4	Partly met
img-53	How could a TROP2 PET scan change how ADCs are used? Expected: It would map antigen expression across all lesions and over time, potentially selecting patients, choosing between TROP2 ADCs, and detecting antigen loss at progression to guide a switch to a different target, as PSMA PET does for Pluvicto. First-in-human tracers exist; no outcome data yet.	Reasoning	3	TROP2 PET, TROP2 PET to choose and sequence TROP2 ADCs, TROP2 PET → TROP2 ADC selection	4/4	All met
tech-54	What are the three parts of an antibody-drug conjugate and what does each do? Expected: An antibody that finds the tumour cell, a linker that holds the payload in the blood and releases it inside the cell, and a payload, a very potent chemotherapy such as a topoisomerase-I inhibitor.	Factual	1	Antibody-drug conjugate (ADC), Payload (ADC), Linker (ADC)	4/4	All met
tech-55	What is the bystander effect in ADCs? Expected: A released, membrane-permeable payload diffuses out of the targeted cell and kills neighbouring cells that lack the antigen; requires a cleavable linker and permeable payload (DXd, MMAE, SN-38), explaining T-DXd's activity in HER2-low disease and T-DM1's lack of it.	Factual	2	Bystander effect (ADC), Trastuzumab deruxtecan, Trastuzumab emtansine	4/4	All met
tech-56	What is drug-to-antibody ratio and why does it matter? Expected: The number of payload molecules per antibody, typically 2-8. Higher DAR delivers more drug per binding event (T-DXd about 8) but increases hydrophobicity and clearance unless hydrophilic linkers are used.	Factual	2	Drug-to-antibody ratio (DAR), Site-specific conjugation & linker chemistry	4/4	All met
tech-57	What is synthetic lethality and what is the proven clinical example? Expected: Two genes where losing either alone is survivable but losing both kills the cell; tumours with BRCA loss die when PARP is inhibited, the basis of olaparib and other PARP inhibitors.	Factual	2	Synthetic lethality, PARP inhibitors, BRCA1 / BRCA2 (HRD)	4/4	All met
tech-58	How does a PROTAC differ from a conventional inhibitor? Expected: Rather than blocking a protein's active site it recruits an E3 ubiquitin ligase to tag the protein for destruction by the proteasome, removing the whole protein including scaffolding functions, catalytically and at sub-stoichiometric doses.	Reasoning	2	PROTACs & molecular glues (targeted protein degradation), Vepdegestrant	1/4	Partly met
tech-59	What is in vivo CAR-T and why is it exciting? Expected: Targeted lipid nanoparticles or viral vectors deliver CAR-encoding mRNA or DNA to T cells inside the patient, avoiding manufacturing, lymphodepletion, and wait time and allowing redosing; first-in-human data in 2025-26 show B-cell depletion. AbbVie bought Capstan for up to $2.1B.	Factual	2	In vivo CAR-T, CAR-T cell therapy	4/4	All met
tech-60	What is a bispecific ADC and why might it beat a normal one? Expected: Its antibody binds two different tumour antigens (for example EGFR and HER3), improving avidity, internalisation, and tumour selectivity and hitting cells that express either antigen, addressing heterogeneity; iza-bren is the first with positive phase 3 data.	Reasoning	2	Bispecific ADC, Izalontamab brengitecan	4/4	All met
tech-61	What is FLASH radiotherapy? Expected: Delivering a full radiation dose in under a second at ultra-high dose rates (>40 Gy/s), which spares normal tissue in animal models; first-in-human trials (FAST-01/02, Varian) show feasibility but the clinical benefit is unproven.	Factual	2	FLASH radiotherapy	4/4	All met
tech-62	What is brachytherapy and where is it essential? Expected: Placing a radioactive source directly inside or next to the tumour; essential in cervical cancer, used in prostate (seeds or HDR), breast, skin, and eye melanoma.	Factual	1	Brachytherapy, Cervical cancer, Prostate cancer	4/4	All met
tech-63	What is the difference between lutetium-177 and actinium-225 as therapeutic isotopes? Expected: 177Lu emits beta particles (range 1-10 mm, crossfire helps bulky heterogeneous tumours, marrow toxicity, 6.7-day half-life); 225Ac emits alpha particles (range 50-100 µm, very high energy, oxygen-independent clustered DNA damage, good for micrometastases, daughter redistribution and supply constraints).	Factual	2	Alpha vs beta emitters, Radioligand therapy (beta emitters), Targeted alpha therapy	4/4	All met
tech-64	Why is actinium-225 supply a problem? Expected: It is made in tiny quantities from a legacy thorium-229 stockpile or accelerators; commercial-scale expansion is a 3-5 year effort (TerraPower's Philadelphia plant aims for a 20-fold increase), so phase 3 alpha trials and future launches are gated by production.	Reasoning	2	TerraPower Isotopes, Targeted alpha therapy, Orano Med, ITM Isotope Technologies Munich	4/4	All met
tech-65	What is tumour treating fields and which cancers is it approved for? Expected: Wearable electrode arrays delivering alternating electric fields that disrupt cell division; approved for glioblastoma, mesothelioma, NSCLC after platinum, and (2026) locally advanced pancreatic cancer.	Factual	1	Tumour treating fields (TTFields), Optune / Optune Pax (TTFields)	4/4	All met
tech-66	What are foundation models in pathology and what have they enabled? Expected: Very large self-supervised models trained on millions of slides (Virchow, UNI, CONCH, Prov-GigaPath) that adapt to many tasks; they predict molecular status from H&E and power FDA-cleared tools such as ArteraAI Prostate and ArteraAI Breast.	Factual	2	Pathology & radiology foundation models, Digital pathology & AI, Artera	4/4	All met
path-67	Explain the PI3K/AKT/mTOR pathway in one paragraph and name approved drugs against it. Expected: Growth receptors activate PI3K (PIK3CA), which produces PIP3 and recruits AKT; PTEN reverses it; AKT releases mTORC1 to drive growth. Drugs: alpelisib and inavolisib (PI3Kα), capivasertib (AKT), everolimus (mTOR), gedatolisib (PI3K/mTOR, 2026).	Factual	2	PI3K / AKT / mTOR, Capivasertib, Inavolisib, Gedatolisib	5/5	All met
path-68	Why does blocking BRAF alone cause problems that adding a MEK inhibitor fixes? Expected: BRAF inhibitor monotherapy causes paradoxical MEK/ERK reactivation through CRAF dimers (and secondary skin cancers); blocking MEK downstream closes the escape and improves PFS and OS in melanoma.	Reasoning	3	BRAF inhibitor + MEK inhibitor, BRAF, RAS / RAF / MEK / ERK (MAPK)	4/4	All met
path-69	What does the cGAS-STING pathway have to do with radiation and ADCs? Expected: DNA damage from radiation, PARP inhibitors, and TOP1 payloads spills DNA into the cytoplasm; cGAS senses it, STING triggers type I interferon and chemokines that recruit and prime T cells, underlying immunogenic cell death, the abscopal effect, and ADC+IO synergy.	Reasoning	3	cGAS-STING innate sensing, Abscopal effect, ADC + checkpoint inhibitor	4/4	All met
path-70	What is oncogene addiction? Expected: When a cancer depends so completely on one mutated gene that blocking it collapses the tumour, as with EGFR, ALK, BCR-ABL, KIT, and BRAF V600E; dramatic responses then acquired resistance.	Factual	1	Oncogene addiction, Small-molecule kinase inhibitors	4/4	All met
who-71	Which hospital is ranked first in the world for oncology in 2026 and by whom? Expected: Memorial Sloan Kettering Cancer Center, by Newsweek/Statista's World's Best Specialized Hospitals 2026; MD Anderson is second.	Factual	1	Memorial Sloan Kettering Cancer Center, MD Anderson Cancer Center, Newsweek/Statista World's Best Specialized Hospitals: Oncology	1/4	Partly met
who-72	How many NCI-designated cancer centers are there and what types? Expected: 74 in 2026: 58 Comprehensive, 8 Clinical, 8 Basic Laboratory.	Factual	1	NCI-Designated Cancer Centers, National Cancer Institute (NIH)	2/4	Partly met
who-73	Where was FAPI PET invented? Expected: Heidelberg (University Hospital, DKFZ/NCT; Haberkorn, Giesel, Kratochwil), which also ran the first-in-human 225Ac-PSMA therapy.	Factual	2	Heidelberg University Hospital / NCT / DKFZ, FAPI PET	1/4	Partly met
who-74	Which company owns Trodelvy and how did it get it? Expected: Gilead Sciences, through its $21 billion acquisition of Immunomedics in 2020.	Factual	1	Gilead Sciences (incl. Kite), Sacituzumab govitecan	0/4	Missed
who-75	Which company develops sacituzumab tirumotecan and who holds ex-China rights? Expected: Sichuan Kelun-Biotech developed sac-TMT (approved in China 2024); Merck holds ex-Greater-China rights and runs the TroFuse phase 3 programme; FDA priority voucher July 2026.	Factual	2	Sacituzumab tirumotecan, Sichuan Kelun-Biotech, Merck & Co. (MSD)	4/4	All met
who-76	Where can I look up whether a specific tumour mutation has an approved drug? Expected: OncoKB (MSK, FDA-recognised levels of evidence) and CIViC (open, community-curated); cBioPortal to see mutation frequency; ClinicalTrials.gov for trials.	Procedural	1	OncoKB, CIViC, cBioPortal for Cancer Genomics, ClinicalTrials.gov	1/4	Partly met
who-77	What is DepMap for? Expected: The Broad Institute's Cancer Dependency Map: genome-wide CRISPR and drug screens across more than 1,000 cancer cell lines showing which genes each line cannot live without, the source of synthetic-lethal targets like PRMT5/MTAP.	Factual	2	DepMap (Cancer Dependency Map), CRISPR functional genomics, Broad Institute of MIT and Harvard	2/4	Partly met
who-78	Which cooperative group ran the trial that put nivolumab into first-line Hodgkin lymphoma? Expected: SWOG (S1826, nivolumab-AVD vs BV-AVD), leading to FDA approval in March 2026 for ages 12 and over.	Factual	2	SWOG Cancer Research Network, Nivolumab, Hodgkin lymphoma	4/4	All met
proc-79	My relative has metastatic pancreatic cancer. How do we find a clinical trial? Expected: Ask the oncologist about trials at the treating centre and referral to an NCI-designated or major academic centre; search ClinicalTrials.gov (status recruiting, condition pancreatic cancer, filter by location); OnCo's cancer page has a live recruiting-trials list; ask about RAS inhibitor trials such as daraxonrasib.	Procedural	1	Pancreatic ductal adenocarcinoma, ClinicalTrials.gov, Daraxonrasib, Cancer Commons	2/4	Partly met
proc-80	What should I ask before starting an ADC such as Enhertu? Expected: How lung inflammation (ILD) will be monitored and what symptoms to report immediately; HER2 status and how it was scored; expected nausea, hair loss, and blood counts; dosing schedule (every three weeks); how response will be assessed and when; what comes next if it stops working.	Procedural	2	Trastuzumab deruxtecan, Interstitial lung disease (ILD) / pneumonitis, HER2-low and HER2-ultralow	4/5	Partly met
proc-81	Who should have germline genetic testing when diagnosed with cancer? Expected: All patients with TNBC, ovarian, pancreatic, or metastatic prostate cancer, increasingly all breast cancer, and anyone with suggestive family history; results change surgery, drug choice (PARP inhibitors), and family screening.	Procedural	2	Germline (hereditary) testing, BRCA1 / BRCA2 (HRD), Triple-negative breast cancer (TNBC), Ovarian cancer, Pancreatic ductal adenocarcinoma, Prostate cancer	5/5	All met
proc-82	How do I get a second opinion for a rare cancer? Expected: Ask your oncologist for referral to a high-volume centre (NCI-designated in the US, ESMO-accredited or national reference centres elsewhere); nonprofits such as Cancer Commons offer free navigation; bring pathology slides, genomic reports, and imaging; telemedicine second opinions exist.	Procedural	1	Cancer Commons, NCI-Designated Cancer Centers, Memorial Sloan Kettering Cancer Center	1/4	Partly met
proc-83	What is exercise oncology and is there real evidence? Expected: Structured exercise during and after treatment; the CHALLENGE trial (2025) randomised colon cancer survivors and showed improved disease-free and overall survival, the first level-1 evidence, plus less fatigue and neuropathy.	Factual	1	Exercise & lifestyle oncology, Colorectal cancer	4/4	All met
proc-84	Does scalp cooling prevent hair loss from chemotherapy? Expected: Partly: roughly half of patients on taxane-based regimens keep at least 50% of their hair with DigniCap or Paxman; it is less effective with anthracyclines and reimbursement varies.	Factual	1	Scalp cooling (cold caps: DigniCap, Paxman)	4/4	All met
proc-85	What does a 'Phase 3' trial mean compared with Phase 1? Expected: Phase 1 tests safety and dose in a small group; Phase 2 looks for activity; Phase 3 randomises hundreds to thousands of patients against the current standard to prove benefit, the usual basis for full approval.	Factual	1	Lines of therapy, Accelerated approval, Standard of care	2/4	Partly met
proc-86	What is accelerated approval and what are its risks for patients? Expected: FDA approval based on early evidence such as tumour shrinkage or pCR, conditional on a confirmatory trial; if confirmation fails the drug is withdrawn (atezolizumab in TNBC, belantamab in 2022, sacituzumab in bladder cancer), so benefit is not yet proven when prescribed.	Reasoning	2	Accelerated approval, IMpassion130, Belantamab mafodotin	4/4	All met
pipe-87	List the generations of ADC technology and what defined each. Expected: First (gemtuzumab, 2000: unstable linker, heterogeneous); second (brentuximab 2011, T-DM1 2013: humanised antibodies, potent tubulin payloads, stable linkers, no bystander for T-DM1); third (T-DXd, sacituzumab, 2019-22: high-DAR TOP1 payloads with bystander effect); fourth emerging (bispecific, dual-payload, degrader, immune-stimulating, masked, radio-ADCs).	Factual	3	ADC roadmap: from Mylotarg to bispecific and dual-payload ADCs, Antibody-drug conjugate (ADC)	3/4	Partly met
pipe-88	What are the three approved or near-approved TROP2 ADCs and how do their payloads differ? Expected: Sacituzumab govitecan (SN-38, DAR ~7.6, hydrolysable linker), datopotamab deruxtecan (DXd, DAR ~4, GGFG cleavable linker), and sacituzumab tirumotecan (belotecan-derived T030, DAR ~7.4, stable linker; approved in China, US decision pending).	Factual	3	Sacituzumab govitecan, Datopotamab deruxtecan, Sacituzumab tirumotecan, TROP2 ADC roadmap: sacituzumab govitecan → Dato-DXd → sac-TMT → bispecifics and PET	4/4	All met
pipe-89	Which oncology drugs were approved in the US in July 2026? Expected: Isatuximab subcutaneous (Sarclisa Escena), pembrolizumab SC plus enfortumab vedotin, selpercatinib label update, gedatolisib (Revtorpyk), zidesamtinib (Jideytro), and a Pluvicto label expansion.	Factual	3	Gedatolisib, Zidesamtinib, Lutetium-177 vipivotide tetraxetan, Selpercatinib, FDA Oncology Approvals (OCE) & Novel Drug Approvals	0/4	Missed
pipe-90	What happened to tazemetostat in 2026? Expected: Ipsen withdrew Tazverik from all markets and indications on 9 March 2026 after SYMPHONY-1 showed excess secondary haematologic malignancies (5.7% vs none), and stopped its trials.	Factual	2	EZH2	2/4	Partly met
pipe-91	What is the readout calendar event on 23 September 2026? Expected: The FDA advisory committee (Molecular and Clinical Genetics Panel) reviewing GRAIL's Galleri multi-cancer early detection PMA.	Factual	1	Galleri, Multi-cancer early detection (MCED)	1/4	Partly met
pipe-92	Name three next-generation ADC concepts beyond bispecific antibodies. Expected: Dual-payload ADCs, degrader-antibody conjugates, immune-stimulating antibody conjugates, masked or conditionally active ADCs, peptide-drug conjugates, radio-ADCs.	Factual	2	Dual-payload ADC, Degrader-antibody conjugate (DAC), Immune-stimulating antibody conjugate (ISAC), Masked / conditionally active ADC	1/4	Partly met
pipe-93	Which company's PD-1×VEGF bispecific did BMS license, and for how much? Expected: BioNTech's pumitamig (BNT327), in a 2025 deal valued at about $11 billion.	Factual	2	BioNTech, Bristol Myers Squibb	4/4	All met
pipe-94	What is the rationale for combining an ADC with a checkpoint inhibitor, and where is it proven? Expected: ADC payloads cause immunogenic cell death and STING activation, releasing antigens and depleting suppressive myeloid cells; PD-1 blockade converts that priming into durable control. Proven in urothelial cancer (EV-302) and TNBC (ASCENT-04).	Reasoning	2	ADC + checkpoint inhibitor, EV-302 / KEYNOTE-A39, ASCENT-04 / KEYNOTE-D19	3/4	Partly met
pipe-95	Why do OnCo's institution rankings differ from research-output rankings? Expected: OnCo's score sums Newsweek 2026 rank points, NCI designation points, and links in the corpus, so it measures clinical reputation plus documentation coverage, while Nature Index and SCImago count publications; both formulas are disclosed and neither measures clinical benefit.	Reasoning	2	Newsweek/Statista World's Best Specialized Hospitals: Oncology, Nature Index: Healthcare and Cancer institutions, SCImago Institutions Rankings: Oncology	4/4	All met
reason-96	A patient with HER2-low, hormone-positive metastatic breast cancer has progressed on a CDK4/6 inhibitor and one chemotherapy. What are the ADC options and how would you order them? Expected: T-DXd (DESTINY-Breast04/06) is generally first among ADCs for HER2-low disease; sacituzumab govitecan (TROPiCS-02) or Dato-DXd (TROPION-Breast01) later. Both classes carry TOP1 payloads, so cross-resistance is a concern and interposing chemotherapy is debated; consider ESR1/PIK3CA-directed options if endocrine-sensitive.	Reasoning	3	Trastuzumab deruxtecan, Sacituzumab govitecan, Datopotamab deruxtecan, ADC sequencing, HR-positive / HER2-negative breast cancer, HR-positive metastatic breast cancer after CDK4/6 inhibitors	4/4	All met
reason-97	Why are mesenchymal or claudin-low triple-negative tumours hard to treat with any class of drug? Expected: EMT programmes (ZEB1/2, SNAIL) confer stemness, apoptosis resistance, immune exclusion, and high ABC efflux transporters that pump out chemotherapy and ADC payloads (MMAE, DXd, SN-38), so efflux-independent approaches (radiation, radioligands, immune killing) are proposed.	Reasoning	3	Epithelial-mesenchymal transition & drug efflux, Drug efflux pumps (ABC transporters), Efflux-agnostic therapy for mesenchymal TNBC, Triple-negative breast cancer (TNBC)	4/4	All met
reason-98	Is a positive ctDNA result after surgery a reason to start treatment right away? Expected: It depends on proof that intervening helps: in bladder cancer IMvigor011 showed atezolizumab improves survival in ctDNA-positive patients (now approved); in colon cancer DYNAMIC supports de-escalation for negatives; in breast cancer trials are ongoing and lead time without a proven intervention mainly causes anxiety.	Reasoning	3	ctDNA MRD → adjuvant therapy decision, IMvigor011, DYNAMIC, ctDNA-triggered escalation in early TNBC	3/4	Partly met
reason-99	Why might personalised vaccines work better after surgery than in metastatic disease? Expected: Tumour burden is low so immune responses are not overwhelmed, there is time for the six-week manufacturing, immunosuppression from bulky disease is absent, and recurrence-free survival is a clean endpoint; INTerpath-001 tested exactly this adjuvant setting.	Reasoning	2	Personalised neoantigen (mRNA) vaccines, INTerpath-001 (V940-001), Personalised neoantigen vaccine + PD-1 blockade	3/4	Partly met
reason-100	What single change would most improve outcomes across all cancers, and what evidence supports it? Expected: Prevention and early detection: HPV vaccination has driven cervical cancer toward elimination in vaccinated cohorts, tobacco control and screening (mammography, colonoscopy, low-dose CT) have proven mortality benefits, and early-stage disease is where surgery cures; MCED blood tests aim to extend this but are unproven.	Reasoning	3	HPV & HBV vaccination, Prevention & Risk, Early Detection & Screening, Multi-cancer early detection (MCED), Early detection roadmap: organ screening → blood tests for many cancers	3/4	Partly met