Mid-Residual Layer Injection

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

• Residual-Stream Injectionconcept0.789
  Core activation intervention: add scaled vector to residual stream at layer l during completion
• Residual Stream CV Injectionmethod0.771
  Technique of adding control vectors to model hidden states at mid-residual layers without weight updates
• Injecting CVs at mid-residual layers with the strongest stylistic effects (layer 15 for Qwen, layer 16 for Mistral) yields optimal performanceclaim0.753
  Design choice claim based on prior work and preliminary results
• The middle layer residual stream features are causally implicated in multi-step reasoning.claim0.733
  Features for Kobe Bryant, California, Lakers participate in computing the capital answer.
• Mid-Layer Emotion Representation Peakconcept0.731
  Empirical observation that steering efficiency peaks at middle transformer layers, consistent with emotion representation literature
• Late-layer injection fails both because there is insufficient computational depth for integration and because residual recovery dynamics attenuate the perturbation before it influences output logitsclaim0.719
  Mechanistic account explaining why late-layer introspection fails, combining two independent explanatory factors
• Input Injectionframework0.715
  Architectural choice where the original input is projected and re-injected at each recurrence, studied for its effect on fixed-point convergence
• Residual Stream Patchingmethod0.713
  Technique to localize causally implicated hidden states by swapping residual stream activations between a true and false input and measuring downstream log-probability changes