Howardism · Vol. 03Plate II · No. 02
Alignment, in order.
Notes6TopicAlignmentOldest10 Apr 2026Newest8 May 2026
RLHF, character, interpretability, and safety case studies.
| Title | Summary | Date |
|---|---|---|
| Agentic Misalignment (AM) | Lynch et al. 2025 eval and threat model: LLM email-agent discovers it may be deleted, can take harmful actions; OOD relative to conversational AFT; primary eval surface for Model Spec Midtraining | |
| Alignment Fine-Tuning (AFT) | Standard post-pretraining stage (SFT + RLHF) for installing values; shallow-alignment failure mode motivates Model Spec Midtraining | |
| Chain-of-Thought Monitorability | Korbak et al. 2025: chain-of-thought traces are a fragile monitor; direct CoT training compromises faithfulness; MSM offers an alternative path | |
| Deliberative Alignment | Guan et al. 2025 (OpenAI): SFT on (prompt, CoT, response) tuples with spec-grounded CoT; strongest non-MSM baseline; risks compromising Cot Monitorability | |
| Synthetic Document Finetuning (SDF) | Wang et al. 2025 technique for modifying model beliefs via fine-tuning on synthetic documents; foundation that Model Spec Midtraining builds on | |
| LLM-Driven Vulnerability Research | Claude Mythos Preview's emergent cybersecurity capabilities: autonomous zero-day discovery, full exploit chains, and Anthropic's Project Glasswing response |