Low-Rank Adaptation (LoRA)

Parameter-efficient fine-tuning method used to implement SOO fine-tuning on LLMs

Neighborhood — ranked by edge-count

framework

Self-Other Overlap (SOO) Fine-Tuning
uses
The central framework proposed in this paper: aligning AI internal representations of self and others to reduce deceptive behavior

method

LoRA (Low-Rank Adaptation)
same_as
Parameter-efficient fine-tuning method used for both SDF and expert iteration stages.

cosine ≥ 0.65 · no typed edge

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

LoRA: Low-Rank Adaptation of Large Language Models (Hu et al., 2022)concept0.890
Fine-tuning method paper whose technique is used in the fine-tuning experiments
LoRA Adaptersmethod0.778
Parameter-efficient fine-tuning method used in both distillation and introspection stages
A LoRA adapter can learn to add a constant vector equivalent to steering along a pre-existing persona direction, making fine-tuning literally equivalent to activation steeringclaim0.756
Mechanistic explanation of how fine-tuning can shift persona vectors without directly updating activations
LoRA adapterconcept0.754
Fine-tuning technique that can effectively add a constant vector to the residual stream, equivalent to steering along a persona direction
LoRA Fine-Tuningmethod0.745
Adaptation method used via Tinker API for DeepSeek-V3.1 and Qwen3-235B fine-tuning with rank 32
Short rationales (LoRA+CoT) sometimes improve in-distribution performance but do not reliably reduce cross-base harmfinding0.727
E2 finding showing CoT's limited benefit for OOD transfer, consistent with larger dr out of scope
LoRA Fine-Tuning with Axolotlmethod0.724
Specific fine-tuning implementation using LoRA rank 32, learning rate 2e-4, AdamW 8-bit optimizer
LoRA+CoTmethod0.711
Fine-tuning with chain-of-thought rationales aiming to reduce dr via procedural alignment.