finding

active

finding:four-features-a-0-20-a-0-0-a-0-30-a-0-494-form-an-fsa-like-system-implementing-html-tag-generation

Four features (A/0/20, A/0/0, A/0/30, A/0/494) form an FSA-like system implementing HTML tag generation

Concrete example of features connecting into FSA-like system implementing complex behavior

Source paper

extracted_from

Towards Safe and Honest AI Agents with Neural Self-Other Overlap

(2024) · Marc Carauleanu · Michael Vaiana · Judd Rosenblatt · Cameron Berg +1

Neighborhood — ranked by edge-count

Concepts (1)

concept

Finite State Automata Feature Assemblies
supports
Collections of features that interact via the token stream — one feature increases probability of tokens that activate the next feature — forming FSA-like systems

Related by similarity (8)

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.

Single base64 feature A/0/45 splits into three distinct features in A/1: letter-specific, digit-specific, and ASCII-encoding-specificfinding0.762
Concrete example of feature splitting revealing unexpected model structure
In A/4, features functionally memorize 'MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE' via FSA-like feature chainfinding0.746
Demonstrates mechanistic memorization via feature assemblies in superposition
Four features will be fundamental: (1) all processes rethought as morphogenetic; (2) sequences link into a net; (3) continuous evolution becomes widespread; (4) an ethical obligation to heal the land emerges.claim0.734
Vision of the emerging paradigm shift in society.
Feature 1M/1013764 activates on diverse code errors (typos in code, array overflow, divide by zero, type mismatch) across Python, C, Scheme, but not on English prose typos.finding0.718
Shows a general code error detector beyond simple typo detection.
In A/4, over 100 features primarily respond to the token 'the' in different contextsfinding0.713
Demonstrates prevalence of token-in-context features and feature splitting of common tokens
The resulting features are highly abstract: multilingual, multimodal, and generalizing between concrete and abstract references.claim0.706
Features respond to concepts across languages and in images, not just text.
Feature ablation (zeroing feature activations)method0.697
Clamping a feature's value to zero to measure its causal effect on model output.
Qwen3.5-9B and Claude Opus 4.6 evolvers produce procedurally isomorphic flink-query skills that both enable Opus 4.6 agent to score 1.0 vs. 0.67 without skillfinding0.696
Case study demonstrating mechanism behind flat harness-updating: smaller models reach same procedural content