Tracing the thoughts of a large language model

linked as “Anthropic said”

linked as “related work from 2025”

linked as “tracing the thoughts of a language model”

linked as “experiments where concepts are injected directly into a model's activations”

linked as “experiments where concepts are injected directly into a model's activations”

linked as “circuit-tracing and attribution-graph methods that let researchers follow how a concept moves from p…”

linked as “how AIs compose poetry”

linked as “Anthropic’s research on tracing thoughts in language models”

linked as “link”

linked as “I keep reading that you guys have no idea how the stuff you built works either”

linked as “in really weird ways”

linked as “looks more like biology research than engineering”

linked as “Tracing the Thoughts of an LLM”

linked as “Tracing the thoughts of a large language model”

linked as “how LLMs might “think””

linked as “can’t explain why an LLM returns a specific result”

linked as “how it performs arithmetic”

linked as “Tracing the thoughts of a large language model \ Anthropic”

linked as “Tracing the thoughts of a large language model”

linked as “a lot of breakthroughs lately”

linked as “Tracing the thoughts of a large language model (8 minute read)”

linked as “recent reports from Anthropic”

linked as “great strides in this area”

linked as “how they actually are doing the math”