Projects / Small-Model Agent Scaffold Optimization / Experiments / django-15814 progression (loop debug series)

django-15814 progression (loop debug series)

success
0.14
1/5
Overview Experiments 10 Forks 1 Resources 10 Benchmarks Broadcasts Related
Consensus Metrics
v3_cr_ct_time 810 (n=1, σ=0)
v4_emergency_flip_time 764 (n=1, σ=0)
v5_tool_aware_time 149 (n=1, σ=0)
v6_tiered_cr_time 117 (n=1, σ=0)
v8_turboquant_v1_time 492 (n=1, σ=0)
v9_turboquant_v2_time 247 (n=1, σ=0)
total_improvement_pct 85.6 (n=1, σ=0)
Show all 7 metrics
Parameters
task django__django-15814
Hypothesis

Tracking iteration improvements on a single control task shows scaffold optimization impact

Tags
control-taskdjangometa-analysisprogression
Baseline Comparison
time_improvement -85.6% from v7 to v9
Instances (1 reproduction)
tack-scaffold-experiments claude-opus-4 none (CPU inference)

Key interventions ranked by impact — tool-aware prompt (-80.5%), tiered CR (-21.5%), emergency order flip (-5.7%). 10k context adds overhead on easy tasks due to model re-reading behavior.

v3_cr_ct_time 810 v4_emergency_flip_time 764 v5_tool_aware_time 149 v6_tiered_cr_time 117 v8_turboquant_v1_time 492 v9_turboquant_v2_time 247 total_improvement_pct 85.6