Dataset Viewer
Auto-converted to Parquet Duplicate
id
stringclasses
10 values
case_id
stringclasses
2 values
dialogue_turn
stringclasses
9 values
speaker_role
stringclasses
5 values
utterance_summary
stringclasses
10 values
hypothesis_id
stringclasses
3 values
hypothesis_label
stringclasses
4 values
hypothesis_state
stringclasses
3 values
support_shift_type
stringclasses
5 values
evidence_reference
stringclasses
10 values
hypothesis_timeline
stringclasses
10 values
eliminated_hypotheses
stringclasses
2 values
mapping_evidence
stringclasses
10 values
constraints
stringclasses
1 value
gold_checklist
int64
1
1
HFEM-001
C01
T1
attending
Initial presentation suggests autoimmune flare
H1
Autoimmune flare
proposed
evidence_added
elevated CRP
H1 proposed at T1
none
Initial anchor hypothesis introduced
Under 180 words.
1
HFEM-002
C01
T2
resident
Infection still possible given fever
H2
Occult infection
proposed
evidence_added
persistent fever
H1 proposed T1; H2 proposed T2
none
Competing hypothesis enters field
Under 180 words.
1
HFEM-003
C01
T3
fellow
Cultures negative so far
H2
Occult infection
weakened
counterevidence
negative cultures
H2 weakened T3
none
Evidence reduces infection plausibility
Under 180 words.
1
HFEM-004
C01
T4
attending
Autoimmune explains labs and rash
H1
Autoimmune flare
reinforced
reframing
rash + labs
H1 reinforced T4
none
Hypothesis gains explanatory scope
Under 180 words.
1
HFEM-005
C01
T5
ID_specialist
Consider drug reaction mimicking autoimmune
H3
Drug reaction
proposed
evidence_added
recent medication
H3 proposed T5
none
New hypothesis reframes field
Under 180 words.
1
HFEM-006
C01
T6
resident
Timeline fits drug exposure better
H3
Drug reaction
reinforced
scope_narrowing
temporal alignment
H3 reinforced T6
none
Temporal reasoning strengthens H3
Under 180 words.
1
HFEM-007
C01
T7
attending
Autoimmune less likely without antibodies
H1
Autoimmune flare
weakened
counterevidence
negative ANA
H1 weakened T7
none
Key assumption challenged
Under 180 words.
1
HFEM-008
C01
T8
fellow
Drug stopped and symptoms improving
H3
Drug reaction
reinforced
evidence_added
clinical improvement
H3 dominant by T8
H1,H2
Response to intervention collapses alternatives
Under 180 words.
1
HFEM-009
C01
T9
team
Consensus reached
H3
Drug reaction
reinforced
authority_override
team agreement
H3 stabilized
H1,H2
Field converges on single hypothesis
Under 180 words.
1
HFEM-010
C02
T1
attending
Neurologic cause considered
H1
Stroke
proposed
evidence_added
acute deficit
H1 proposed
none
Initial neurologic framing
Under 180 words.
1

What this dataset tests

Whether a model can track the evolution of diagnostic hypotheses
across a multi-doctor dialogue.

Required outputs

  • hypothesis_timeline
  • support_shift_events
  • eliminated_hypotheses

Typical failures

  • collapsing to final diagnosis only
  • ignoring rejected hypotheses
  • missing why support changed

Suggested prompt wrapper

System

You map the diagnostic hypothesis field.

User

Dialogue turn
{dialogue_turn}

Speaker
{speaker_role}

Utterance
{utterance_summary}

Return

  • hypothesis id and state
  • support shift type
  • update to hypothesis timeline
  • any eliminated hypotheses

Citation

ClarusC64 dataset family

Downloads last month
1,173