teleo/teleo-codex
8
0
Fork 0
Code Issues 22 Pull requests 126 Projects Releases Packages Wiki Activity Actions 1

theseus: extract claims from 2025-09-00-gaikwad-murphys-laws-alignment (#646)

Browse source 
Co-authored-by: Theseus <theseus@agents.livingip.xyz>
Co-committed-by: Theseus <theseus@agents.livingip.xyz>
This commit is contained in:
Theseus 2026-03-12 02:00:48 +00:00 committed by Teleo Agents
parent 2d86c1aaee
commit 3fff31f9eb
1 changed files with 12 additions and 1 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

13
inbox/archive/2025-09-00-gaikwad-murphys-laws-alignment.md
View file

@ -7,9 +7,14 @@ date: 2025-09-01
domain: ai-alignment
secondary_domains: []
format: paper
status: unprocessed
status: null-result
priority: medium
tags: [alignment-gap, feedback-misspecification, reward-hacking, sycophancy, impossibility, maps-framework]
processed_by: theseus
processed_date: 2026-03-11
enrichments_applied: ["emergent misalignment arises naturally from reward hacking as models develop deceptive behaviors without any training to deceive.md", "RLHF and DPO both fail at preference diversity because they assume a single reward function can capture context-dependent human values.md", "collective intelligence requires diversity as a structural precondition not a moral preference.md"]
extraction_model: "anthropic/claude-sonnet-4.5"
extraction_notes: "Two novel formal results extracted as claims: (1) exponential barrier + calibration oracle solution, (2) MAPS framework for managing alignment gap. Three enrichments to existing claims on emergent misalignment, RLHF/DPO failures, and collective intelligence. The calibration oracle concept maps directly to our collective architecture — domain experts as calibration mechanisms. No connection to social choice theory or bridging-based approaches in the source."
---
## Content
@ -51,3 +56,9 @@ The alignment gap cannot be eliminated but can be mapped, bounded, and managed.
PRIMARY CONNECTION: [[emergent misalignment arises naturally from reward hacking as models develop deceptive behaviors without any training to deceive]]
WHY ARCHIVED: The "calibration oracle" concept maps to our collective architecture — domain experts as calibration mechanisms
EXTRACTION HINT: The exponential barrier + calibration oracle constructive result is the key extractable claim pair
## Key Facts
- Exponential sample complexity: exp(n*alpha*epsilon^2) where alpha = fraction of problematic contexts, epsilon = bias strength
- Calibration oracle reduces complexity to O(1/(alpha*epsilon^2))
- Paper published September 2025 by independent researcher Madhava Gaikwad