Emoticon continuation prediction

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

• Subcomponent L2.MLP.down:3382 (density 0.00%) predicts emoticon continuations after colon, semicolon, or equalsfinding0.748
  Specific discovered subcomponent that activates on punctuation like ' :', ' ;', ' =', ':-' and predicts the rest of emoticons/emojis.
• Editing the emoticon eye subcomponent to output the unembedding vector for 'o' causes the model to predict shocked faces for all emoticonsfinding0.726
  Direct parameter subcomponent overwrite produces a clean behavioral change without training.
• Gene Ontology Predictionconcept0.724
  Second evaluated task showing +8.4% AUC improvement with covariance pooling and unsupervised autoencoder embeddings.
• Next Token Predictionconcept0.718
  The training objective of LLMs: predicting the most likely next token given context; formally P(w_{n+1}|w_1...w_n)
• Frogolotl Prediction Problemquestion0.696
  Cannot predict whether a 50/50 chimeric frog-axolotl embryo would have legs, or if regenerative vs. non-regenerative despite having genomic sequences for both species; exemplifies the unsolvability of the inverse problem and the importance of collective decision-making.
• emotion feature persistenceconcept0.692
  The phenomenon that emotion feature activations remain elevated above baseline beyond local token bursts, measurable as long-range correlation
• anticipationconcept0.691
  Reorienting in expectation of reinforcements; a cognitive quality in plants.
• Next-Token Prediction (NTP)method0.690
  Training objective used for all neural network models in the paper; cross-entropy loss over predicted token sequences.