Research

Towards monosemanticity

Anthropic

Using dictionary learning to decompose a transformer layer into thousands of monosemantic features, a path past the polysemantic-neuron problem.

Visit Site

Research Anthropic

Towards monosemanticity
ResearchNatural Language AutoencodersAnthropic ResearchHow Claude is accelerating protein design and analytical chemistryAnthropic ResearchRed teaming language models to reduce harmsAnthropic ResearchClaude is competitive with humans in (some) cyber competitionsAnthropic ResearchTraining a helpful and harmless assistant with RLHFAnthropic ResearchDistributed representations: Composition and superpositionAnthropic BlogsAnnouncing Netlify Log Drains for DatadogNetlify BlogsRefactoring towards Pure Components in ReactNetlify ResearchRepetition In Repetition Out: Towards Understanding Neural Text Degeneration from the DataCohere ResearchThe IOL-AI Challenge: An Open Challenge towards Advancing Linguistic ReasoningCohere BlogsTemporal Workflows at scale with PlanetScale: Part 1Planetscale BlogsReplit + Chroma: AI for the next billion softwareReplit BlogsBuilding Towards a Holistic Development ServiceReplit ResearchTowards understanding sycophancy in language modelsAnthropic BlogsDefining the database maturity modelPlanetscale ResourcesFlashingpine64.org Products & ServicesCircuit Playground Bluefruit - Bluetooth® Low Energy : Adafruit Industries, Unique & fun DIYadafruit.com Products & ServicesCircuit Playground Express for 4-H : Adafruit Industries, Unique & fun DIY electronics and kitsadafruit.com Products & ServicesKano Computer Kit Complete : Adafruit Industries, Unique & fun DIY electronics and kitsadafruit.com Products & ServicesCircuit Playground Express : Adafruit Industries, Unique & fun DIY electronics and kitsadafruit.com