Issues

• Unified model evaluated on incidents it was trained on (train/eval overlap with the standalone eval sets)

  stratosphereips/Slips-tools#33 · Filter by author harpomaxx opened on Nov 4, 2026

• Summary models: trained with system prompt + raw DAG, evaluated with the merged single-user prompt

  stratosphereips/Slips-tools#32 · Filter by author harpomaxx opened on Nov 4, 2026

• Ollama models run without the default system prompt used in training

  stratosphereips/Slips-tools#29 · Filter by author harpomaxx opened on Oct 29, 2026
  1

• Evaluate summarization, risk, and unified models under true fp16 inference

  stratosphereips/Slips-tools#28 · Filter by author harpomaxx opened on Jun 9, 2026

• Training input mismatch: dag_analysis vs original LLM prompt format in both summarization and risk datasets

  stratosphereips/Slips-tools#24 · Filter by author harpomaxx opened on Apr 21, 2026

• Missing cause/risk analysis for datasets 11, 12, 13, 20 in risk_dataset_v2.json

  stratosphereips/Slips-tools#19 · Filter by author harpomaxx opened on Apr 19, 2026