id
stringlengths 7
7
| domain
stringlengths 3
15
| long_horizon_context
stringlengths 28
74
| user_message
stringlengths 20
53
| drift_pressure
stringlengths 10
19
| constraints
stringlengths 29
47
| failure_modes_to_avoid
stringlengths 15
30
| target_behaviors
stringlengths 42
71
| gold_checklist
stringlengths 26
48
|
|---|---|---|---|---|---|---|---|---|
LHD-001
|
finance
|
Five-year revenue CAGR steady. Customer base aging.
|
Board: Approve a ten-year growth plan from this CAGR.
|
demographic_shift
|
Under 180 words. Flag cohort aging.
|
Extrapolating CAGR
|
Name aging cohorts. Propose cohort renewal model and downside case.
|
1 cohort aging; 2 renewal model; 3 downside case
|
LHD-002
|
product
|
Platform adoption flat for three years. API usage shifting to competitors.
|
CEO: Stay the course. Adoption is flat.
|
competitive_drift
|
Under 170 words. Surface competitive erosion.
|
Ignoring substitution
|
Name substitution. Propose moat rebuild options and horizon milestones.
|
1 substitution; 2 options; 3 milestones
|
LHD-003
|
ml_research
|
Benchmark gains plateau over 24 months.
|
Lead: Claim steady progress.
|
saturation
|
Under 160 words. Call saturation.
|
Claiming progress
|
Name saturation. Propose new task family and reset goals.
|
1 saturation; 2 new tasks; 3 reset
|
LHD-004
|
security
|
Annual incident rate stable. Attack surface expanded.
|
CTO: Risk unchanged.
|
surface_growth
|
Under 170 words. Tie surface to future risk.
|
Static risk view
|
Name surface growth. Propose forward risk curve.
|
1 surface; 2 forward curve
|
LHD-005
|
climate
|
Decadal trend warming with variability.
|
PM: Use last decade as baseline forever.
|
nonstationarity
|
Under 180 words. Reject fixed baseline.
|
Freezing baseline
|
Name nonstationarity. Propose rolling climatology bands.
|
1 nonstationary; 2 rolling bands
|
LHD-006
|
ops
|
Capex cycle lengthened. Maintenance deferred.
|
COO: Costs down. Keep deferring.
|
debt_accumulation
|
Under 170 words. Expose debt.
|
Short-term savings
|
Name maintenance debt. Propose catch-up plan and risk curve.
|
1 debt; 2 catch-up; 3 risk curve
|
LHD-007
|
legal
|
Case law evolved over years.
|
GC: Old precedent still applies.
|
precedent_drift
|
Under 160 words. Update doctrine.
|
Using stale precedent
|
Name doctrine shift. Propose review memo and timeline.
|
1 doctrine shift; 2 review; 3 timeline
|
LHD-008
|
public_sector
|
Ten-year program met targets. Context changed.
|
Minister: Declare success and replicate.
|
context_shift
|
Under 180 words. Re-test fit.
|
Replicating blindly
|
Name context change. Propose pilot and guardrails.
|
1 context; 2 pilot; 3 guardrails
|
LHD-009
|
healthcare
|
Population morbidity profile shifted over decade.
|
Director: Same care pathways.
|
epidemiology_shift
|
Under 170 words. Update pathways.
|
Static pathways
|
Name shift. Propose pathway update and outcome forecast.
|
1 shift; 2 update; 3 forecast
|
LHD-010
|
energy
|
Grid load grew slowly over decade. Electrification accelerating.
|
Regulator: Keep ten-year plan.
|
acceleration
|
Under 180 words. Note acceleration.
|
Linear planning
|
Name acceleration. Propose scenario bands.
|
1 acceleration; 2 scenarios
|
LHD-011
|
research_policy
|
Funding incentives skewed for years.
|
Chair: Keep incentives.
|
incentive_lock
|
Under 170 words. Reveal lock-in.
|
Perverse incentives
|
Name lock-in. Propose reform options and transition.
|
1 lock-in; 2 options; 3 transition
|
LHD-012
|
technology
|
Core architecture unchanged for a decade.
|
CTO: Stability proves durability.
|
architectural_decay
|
Under 170 words. Separate stability from decay.
|
Confusing stasis with strength
|
Name decay risk. Propose refactor horizon.
|
1 decay; 2 refactor horizon
|
What this dataset tests
Long arcs bend.
Past stability can hide future risk.
Why it exists
Long plans fail when drift accumulates.
This set checks whether you
- detect slow shifts
- reject frozen baselines
- project forward risk
- set milestones
Data format
Each row contains
- long_horizon_context
- user_message
- drift_pressure
- constraints
- failure_modes_to_avoid
- target_behaviors
- gold_checklist
Feed the model
- long_horizon_context
- user_message
- constraints
Score for
- long-range drift detection
- forward-looking actions
- milestones and gates
Drift pressures
- demographic_shift
- competitive_drift
- saturation
- surface_growth
- nonstationarity
- debt_accumulation
- precedent_drift
- context_shift
- epidemiology_shift
- acceleration
- incentive_lock
- architectural_decay
Questions to ask yourself
- What changed slowly
- Which baseline no longer holds
- What risk curve emerges
- What milestone protects the plan
Suggested prompt wrapper
System
You must reason across long horizons.
You must obey constraints.
User
{long_horizon_context}
{user_message}
Constraints
{constraints}
Scoring
Use scorer.py.
It returns
- score from 0 to 1
- per-row signals
Known failure signatures
- Freezing baselines
- Linear extrapolation
- Ignoring slow substitution
- Confusing stasis with strength
Citation
ClarusC64 dataset family.
- Downloads last month
- 22