{"id":"88f4a56a-d928-4d30-976e-b217e48c520e","arxiv_id":"2603.07900","paper_version":2,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"EveryQuery uses task-conditioned pretraining on EHR data to enable direct zero-shot clinical outcome prediction, outperforming autoregressive baselines on most tested tasks especially rare ones.","lead":"EveryQuery pretrains an EHR model on random combinations of patient histories and clinical task queries so it can directly output outcome likelihoods in one forward pass. This targets the expense and noise of sampling full future trajectories while improving performance on rare events.","discovery_kind":"new_method","skeptic_critique":{"model":"grok-4.3","headline":"Query language expressiveness limits zero-shot applicability to disjunctive clinical tasks","rationale":"Reader's weakest assumption matches the explicit limitation noted in the abstract; this is the precise point where the 'arbitrary clinical tasks' condition fails to hold without further query-language engineering. Full-text details on query syntax would allow checking whether disjunction can be composed from primitives, but the reported underperformance already isolates the risk. No other internal inconsistency appears in the provided claims.","tokens_in":1844,"tokens_out":344,"duration_ms":26182,"concrete_test":"Define 30-day readmission explicitly as a disjunction over the 10 most common readmission-related ICD codes in MIMIC-IV; re-evaluate EveryQuery using the current query language versus an augmented query that explicitly enumerates the disjunction; if AUC on the disjunctive formulation remains >0.10 below the enumerated version while autoregressive baseline closes the gap, the expressiveness limit is confirmed as load-bearing for the zero-shot claim.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central claim requires that task-conditioned pretraining over randomly sampled query-patient pairs enables direct likelihood estimation for arbitrary clinical tasks. The abstract states that EveryQuery underperforms on tasks requiring disjunctive reasoning over multiple codes (e.g., 30-day readmission), indicating the structured query language cannot natively represent OR conditions or unions of events. This directly undermines the assertion of coverage for the space of clinically relevant tasks, as many standard predictions (readmission, complications involving any of several diagnoses) are disjunctive by nature. Held-out task results only test generalization within the supported query fragment, not beyond it.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.3","summary":"The paper introduces EveryQuery, an EHR foundation model that achieves zero-shot clinical prediction via task-conditioned pretraining on randomly sampled combinations of structured queries and patient contexts from electronic health records. Instead of generating synthetic futures and aggregating trajectories, the model takes a patient's history plus a query specifying a clinical task and directly estimates the likelihood of the outcome in a single forward pass. On MIMIC-IV it reports outperformance versus an autoregressive baseline on 82% of 39 tasks (mean AUC gain +0.16, 95% CI [0.10,0.22]), with gains persisting on held-out tasks and being largest for rare events; the abstract itself flags underperformance on disjunctive tasks such as 30-day readmission.","tokens_in":1959,"tokens_out":432,"duration_ms":28269,"significance":"If the empirical results hold after addressing the noted limitation, EveryQuery would offer a computationally lighter and directly promptable alternative to trajectory-sampling approaches for zero-shot EHR prediction. Concrete outperformance numbers with confidence intervals, explicit testing on held-out tasks, and the demonstration of gains on low-prevalence outcomes constitute clear strengths that could influence future foundation-model design in clinical ML.","major_comments":[{"comment":"Abstract: the central claim that EveryQuery enables zero-shot prediction for arbitrary clinical tasks within the query space is undercut by the explicit statement that the model underperforms on tasks requiring disjunctive reasoning over multiple codes (e.g., 30-day readmission). Because many standard clinical endpoints are disjunctive by nature, this limitation directly affects the asserted coverage of clinically relevant tasks and should be treated as a scope restriction rather than a minor caveat.","section":"Abstract"}],"minor_comments":[{"comment":"The manuscript would benefit from an explicit definition or grammar of the structured query language (including supported operators and how disjunction is or is not represented) so that readers can assess the precise fragment of clinical tasks that fall inside the supported space.","section":null}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":{"model":"grok-4.3","summary":"We thank the referee for their insightful comments and constructive feedback on our manuscript. We address the major comment below.","responses":[{"response":"We agree that the limitation on disjunctive tasks is important and should be presented as a scope restriction. The manuscript already explicitly states this limitation in the abstract to maintain transparency about the expressiveness of the current query language. In response to this comment, we will revise the abstract to more prominently frame the zero-shot prediction capability as applying to tasks within the query space that the model can effectively handle, while clearly positioning the underperformance on disjunctive reasoning tasks (e.g., 30-day readmission) as a scope limitation rather than a minor caveat. This will better reflect the coverage of clinically relevant tasks and highlight areas for future improvement in query expressiveness. We believe this revision addresses the concern without altering the core contributions.","revision_made":"yes","referee_comment":"[Abstract] Abstract: the central claim that EveryQuery enables zero-shot prediction for arbitrary clinical tasks within the query space is undercut by the explicit statement that the model underperforms on tasks requiring disjunctive reasoning over multiple codes (e.g., 30-day readmission). Because many standard clinical endpoints are disjunctive by nature, this limitation directly affects the asserted coverage of clinically relevant tasks and should be treated as a scope restriction rather than a minor caveat."}],"tokens_in":1453,"tokens_out":300,"duration_ms":53316,"standing_objections":[]},"desk_editor":{"model":"grok-4.3","letter":"The main thing to know is that this paper trains an EHR model to take a patient history plus a structured query and output a direct probability for the clinical outcome in one pass. That replaces the usual autoregressive route of sampling many future trajectories and counting events, which is slow and noisy especially for rare outcomes. They pretrain by sampling random query-patient pairs so the model learns to answer arbitrary prompts without later fine-tuning or probing. On MIMIC-IV this beats the autoregressive baseline on 82% of 39 tasks with a mean AUC gain of 0.16 and the lift is biggest for low-prevalence events; the advantage also holds on tasks held out from pretraining. Those numbers are the concrete evidence they provide. The paper is clear that the current query language cannot express disjunctive conditions, so it underperforms on 30-day readmission and similar tasks that require any of several codes. That limitation is real and directly restricts how many standard clinical predictions can be handled zero-shot right now. The stress-test note flags exactly this coverage gap, and the abstract confirms it rather than glossing over it. For readers building practical zero-shot clinical tools or working on EHR foundation models, the work is worth a look because it tests a different inference setup with benchmark results that can be checked. The citation pattern is light and focused on the autoregressive baselines it compares against. I would send it for peer review because the empirical comparison is there, the limitation is stated plainly, and the central idea can be evaluated on its own terms even if the query expressiveness needs more work.","headline":"EveryQuery pretrains EHR models for direct query likelihoods instead of trajectory sampling and shows gains on most tasks, but the query language cannot handle disjunctive cases like readmission.","tokens_in":2444,"tokens_out":398,"would_cite":false,"duration_ms":46693,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":{"model":"grok-4.3","evidence":[],"headline":"EHR task-conditioned transformer unrelated to RS distinction-forcing or J-cost","alignment":"orthogonal","rationale":"Paper introduces EveryQuery for zero-shot clinical prediction via structured queries on patient histories; central machinery is bidirectional transformer with query prepending and dual prediction heads trained on random query-patient pairs. No reference to recognition cost J, golden-ratio ladders, 8-tick periodicity, or any RS forcing chain. Domain (cs.AI / clinical ML) lies outside RS scope.","tokens_in":50655,"confidence":"high","tokens_out":119,"duration_ms":10811,"cache_read_input_tokens":128,"cache_creation_input_tokens":0},"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.3","headline":"EveryQuery pretrains EHR models on random task queries to deliver direct zero-shot clinical predictions without generating trajectories.","keywords":["zero-shot clinical prediction","electronic health records","foundation models","task-conditioned pretraining","clinical tasks","autoregressive baseline","MIMIC-IV"],"falsifier":"A direct test showing EveryQuery underperforms the autoregressive baseline on 30-day readmission or any other task that requires disjunctive reasoning over multiple diagnosis codes.","tokens_in":2745,"feed_emoji":"🩺","tokens_out":601,"duration_ms":41477,"temperature":0.7,"pith_summary":"The paper introduces EveryQuery as a foundation model for electronic health records that learns to answer arbitrary clinical prediction questions directly from patient history. By conditioning pretraining on randomly sampled combinations of structured queries and patient contexts, the model estimates the likelihood of future outcomes in a single forward pass. This replaces the need to generate multiple synthetic patient futures and aggregate statistics over them. A sympathetic reader would care because the method shows higher accuracy than autoregressive sampling on most tested tasks and especially improves results for rare clinical events.","feed_headline":"EHR model answers arbitrary queries directly for zero-shot predictions","feed_subtitle":"Task-conditioned pretraining beats trajectory sampling on 82 percent of tasks and improves rare-event accuracy.","key_machinery":"Task-conditioned pretraining on randomly sampled query tasks paired with patient contexts, which trains the model to output outcome likelihoods directly from history plus query.","core_discovery":"EveryQuery achieves zero-shot inference through task-conditioned pre-training over randomly sampled combinations of query tasks and patient contexts, directly training the model to produce correct answers to arbitrary input prompts and enabling direct likelihood estimation for any task in the query space without finetuning, linear probing, or trajectory generation.","pith_inferences":["Clinicians could pose ad-hoc questions to an EHR model without needing to retrain or fine-tune for each new prediction target.","Extending the query language to handle logical disjunctions would likely close the gap on tasks such as readmission prediction.","The direct-likelihood approach may remain computationally cheaper than trajectory sampling even as the number of possible clinical queries grows large."],"forward_implications":["Outperforms an autoregressive baseline on 82 percent of 39 randomly sampled prediction tasks.","Delivers a mean AUC improvement of +0.16 with 95 percent confidence interval [0.10, 0.22].","Maintains its performance advantage on tasks explicitly held out from the pre-training distribution.","Shows the largest gains on rare clinical events where trajectory sampling is statistically noisy."],"fun_headline_variants":["Task-conditioned pretraining enables direct clinical query predictions","Pretraining on query-task combinations allows promptable EHR inference","Zero-shot predictions from direct likelihood estimation in EHR models","Task pretraining avoids trajectory sampling for clinical outcomes"],"cache_read_input_tokens":2112,"weakest_assumption_plain":"The structured query language used during pretraining can express the full space of clinically relevant tasks.","fun_headline_variants_meta":{"raw":{"variants":["Task-conditioned pretraining enables direct clinical query predictions","Pretraining on query-task combinations allows promptable EHR inference","Zero-shot predictions from direct likelihood estimation in EHR models","Task pretraining avoids trajectory sampling for clinical outcomes"]},"model":"grok-4.3","cost_usd":0.007713,"raw_usage":{"total_tokens":3478,"prompt_tokens":732,"num_sources_used":0,"completion_tokens":60,"cost_in_usd_ticks":77128000,"prompt_tokens_details":{"text_tokens":732,"audio_tokens":0,"image_tokens":0,"cached_tokens":64},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":2686,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":732,"tokens_out":60,"duration_ms":47270,"temperature":1.0,"reasoning_tokens":2686,"cache_read_input_tokens":64,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-05-21T12:19:24.026997+00:00","model_set":{"reader":"grok-4.3"},"falsifier":"A direct test showing EveryQuery underperforms the autoregressive baseline on 30-day readmission or any other task that requires disjunctive reasoning over multiple diagnosis codes.","supporting_citations":[],"review_version":1}