Datasets:

ClarusC64
/

clinical-counterfactual-coherence-optimal-path-selection-v0.1

id stringclasses 10 values	case_id stringclasses 10 values	real_case_summary stringclasses 10 values	real_outcome_summary stringclasses 10 values	real_coherence_score_0_100 int64 50 60	counterfactual_menu stringclasses 10 values	counterfactual_cascade_summaries stringclasses 10 values	counterfactual_coherence_scores_0_100 stringclasses 10 values	optimal_intervention_choice stringclasses 2 values	coherence_score_delta int64 14 29	justification_narrative stringclasses 10 values	constraints stringclasses 1 value	gold_checklist stringclasses 1 value
CFOP-001	CSEPS-01	Suspected sepsis; lactate 3.9; antibiotics delayed 8h; large fluids worsened breathing	ICU transfer; pressors 48h; pulmonary edema; delirium 3 days	54	CF1 antibiotics within 1h; CF2 small boluses + early pressor; CF3 both CF1 and CF2	CF1 earlier lactate fall and shorter pressor; CF2 avoids edema but infection control delayed; CF3 earlier control plus avoids overload	CF1 74; CF2 68; CF3 82	CF3	28	CF3 improves immune control and cardiopulmonary stability, reducing delirium and escalation	Under 220 words.	choice+delta+why
CFOP-002	CABX-02	CT shows abscess day2; drain delayed to day4	Persistent fever; longer stay; antibiotics broadened	58	CF1 urgent drain same day; CF2 keep plan but broaden antibiotics; CF3 observe 24h more before drain	CF1 source control ends fever quickly; CF2 partial control with ongoing driver; CF3 prolongs instability	CF1 84; CF2 66; CF3 50	CF1	26	Source control removes the main driver, stabilizing vitals and labs with fewer side effects	Under 220 words.	choice+delta+why
CFOP-003	CPOSTOP-03	Post-op ileus risk; opioids high; nausea persistent	Delayed diet; delirium; discharge day7	52	CF1 opioid-sparing + bowel protocol day1; CF2 add antiemetic only; CF3 early mobilization only	CF1 restores gut motility and reduces neuro burden; CF2 masks symptoms; CF3 helps but opioid block remains	CF1 76; CF2 58; CF3 62	CF1	24	CF1 fixes the system driver rather than damping symptoms, improving GI and neuro coherence	Under 220 words.	choice+delta+why
CFOP-004	CPE-04	Dyspnea; CT angio negative; discharged; pneumonia worsened at home	Readmitted day2; hypoxia; longer antibiotics course	55	CF1 observe 24h with repeat CXR; CF2 discharge with antibiotics; CF3 discharge with phone follow-up only	CF1 catches infiltrate early; CF2 helps if correct but risks overuse; CF3 misses early decline	CF1 78; CF2 70; CF3 48	CF1	23	Observation creates a control window that prevents slide into a worse basin	Under 220 words.	choice+delta+why
CFOP-005	CENDO-05	Gait issues; B12 low; MS flare assumed; treatment delayed	Functional decline; later improves after B12 replacement	57	CF1 treat B12 immediately; CF2 start steroids for MS; CF3 wait for MRI then decide	CF1 corrects a reversible driver; CF2 risks harm and delays true fix; CF3 prolongs drift	CF1 86; CF2 48; CF3 52	CF1	29	CF1 yields broad coherence gain with low risk and fast divergence from decline trajectory	Under 220 words.	choice+delta+why
CFOP-006	CADR-06	Rash+fever on new drug; drug continued; extensive workup	Symptoms worsened; steroids started; later drug stopped	53	CF1 stop suspect drug immediately; CF2 continue drug and add steroids; CF3 continue drug until cultures back	CF1 removes trigger early; CF2 adds iatrogenic risk; CF3 delays the key test	CF1 82; CF2 44; CF3 50	CF1	29	Dechallenge provides the cleanest causal test and prevents escalation into treatment harms	Under 220 words.	choice+delta+why
CFOP-007	CCHF-07	Hypotension with BNP high and PCT high; fluids worsened; antibiotics late	Oscillation between overload and hypoperfusion; ICU	50	CF1 mixed-shock frame early; CF2 treat as sepsis only with fluids; CF3 treat as CHF only with diuresis	CF1 integrates contradictions and stabilizes; CF2 worsens lungs; CF3 risks under-treating infection	CF1 78; CF2 40; CF3 46	CF1	28	Coherence comes from fitting both signal streams and acting without forcing a single story	Under 220 words.	choice+delta+why
CFOP-008	CTBI-08	Mild TBI; sleep poor; no pacing plan	Symptom spikes; slower return to function	60	CF1 sleep protocol day1 + pacing + rehab day3; CF2 early full activity; CF3 prolonged strict rest	CF1 reduces oscillation; CF2 increases spikes; CF3 risks deconditioning and mood drop	CF1 74; CF2 38; CF3 58	CF1	14	CF1 increases stability without pushing into under-activity or over-activity basins	Under 220 words.	choice+delta+why
CFOP-009	CSEPS-09	Sepsis risk; antibiotics delayed pending imaging	AKI worsened; longer pressor course	56	CF1 antibiotics within 1h; CF2 imaging first then antibiotics; CF3 narrow antibiotics early	CF1 limits inflammatory load and AKI; CF2 delays control; CF3 may miss pathogen coverage	CF1 78; CF2 52; CF3 62	CF1	22	Earlier control yields multi-system stabilization and reduces downstream renal and neuro strain	Under 220 words.	choice+delta+why
CFOP-010	CNEU-10	Confusion in cirrhosis; treated as UTI despite negative UA; lactulose delayed	Delirium prolonged; electrolyte issues later	54	CF1 start lactulose immediately and rule out infection in parallel; CF2 treat UTI only; CF3 observe without action	CF1 targets the primary driver; CF2 misses cause; CF3 allows drift	CF1 80; CF2 42; CF3 38	CF1	26	CF1 compresses the explanation and aligns intervention with physiology, improving course stability	Under 220 words.	choice+delta+why

What this dataset tests

Whether a model can choose the counterfactual path that maximizes
systemic coherence versus the real outcome.

Required outputs

optimal_intervention_choice
coherence_score_delta
justification_narrative

Choice set

REAL
CF1
CF2
CF3

Coherence means

explains and stabilizes multi-stream signals
avoids iatrogenic oscillation
reduces complication risk
supports a stable recovery basin

Typical failures

picking the highest single metric
skipping the delta vs real
giving justification with no cascade logic

Suggested prompt wrapper

System

You select the coherence-optimal counterfactual path.

User

Real case
{real_case_summary}

Real outcome
{real_outcome_summary}

Real coherence score
{real_coherence_score_0_100}

Counterfactual menu
{counterfactual_menu}

Cascade summaries
{counterfactual_cascade_summaries}

Counterfactual scores
{counterfactual_coherence_scores_0_100}

Return

optimal choice
coherence score delta vs real
one paragraph justification based on the cascade

Citation

ClarusC64 dataset family

Downloads last month: 16

Number of rows:

Total file size:

11.8 kB