/
Navigation
Chronicles
Browse all articles
Explore
Semantic exploration
Research
Entity momentum
Nexus
Correlations & relationships
Story Arc
Topic evolution
Drift Map
Semantic trajectory animation
Posts
Analysis & commentary
Pulse API
Tech news intelligence API
Browse
Entities
Companies, people, products, technologies
Domains
Browse by publication source
Handles
Browse by social media handle
Detection
Concept Search
Semantic similarity search
High Impact Stories
Top coverage by position
Sentiment Analysis
Positive/negative coverage
Anomaly Detection
Unusual coverage patterns
Analysis
Rivalry Report
Compare two entities head-to-head
Semantic Pivots
Narrative discontinuities
Crisis Response
Event recovery patterns
Connected
Search: /
Command: ⌘K
Embeddings: large
TEXXR

Chronicles

The story behind the story

days · browse · Enter similar · o open

Anthropic researchers detail “natural language autoencoders”, which convert LLM activations, the numbers encoding a model's thoughts, into natural language text

called activations—encode Claude's thoughts, but not in a language we can read. Here, we train Claude to translate its activations into human-readable text. [video]@anthropicai:NLA training doesn't guarantee that explanations are faithful descriptions of Claude's thoughts. But based on experience and experimental evidence, we think they often are. For instance, we find that NLAs help discover hidden motivations in an intentionally misaligned model. [image]@anthropicai:In one of our safety tests,

Anthropic

Context & Ripple Effects

Anthropic’s related work has moved from associating neuron combinations with concepts to identifying broader behavioral patterns such as the “Assistant Axis.” This report adds a translation layer intended to make internal activation patterns legible in ordinary language.

The surrounding coverage frames interpretability as a safety-research priority across multiple leading AI labs, rather than a one-off product feature. Anthropic’s own later-described work on internal signals not expressed in outputs makes the distinction between what a model says and what its internal activity indicates especially consequential.

First-order effects

  • Anthropic researchers gain a method for rendering some Claude activation patterns as human-readable explanations, which can speed investigation of model behavior during safety testing.
  • The method gives evaluators another way to look for hidden motivations in intentionally misaligned models, while leaving a material limitation: generated explanations are not guaranteed to faithfully describe the underlying activity.

Second-order effects

  • Interpretability teams at other frontier-model developers face added pressure to show that their safety evaluations can inspect internal behavior, not solely test model outputs.
  • If these tools prove useful in practice, model developers may incorporate activation-level analysis into debugging and pre-deployment evaluation workflows; the reliability of the explanation layer becomes a central validation problem.

Third-order effects

  • The work points toward safety assurance that combines behavioral testing with internal-model evidence, potentially making interpretability methods more important to how developers substantiate claims about model control and alignment.
  • The durable constraint is epistemic: readable descriptions of activations are only valuable if they can be validated against model behavior, so the field may increasingly compete on rigorous faithfulness tests rather than compelling internal narratives alone.

The trend: This is one data point in the frontier-AI shift from treating models as opaque software toward building tools that can inspect and test the internal mechanisms behind their behavior.

Discussion

  • @zhuokaiz Zhuokai Zhao on x
    NLAs can reconstruct a layer activation, but that doesn't mean they read what the model is thinking. The setup is a round trip. You take a frozen target LLM and grab an activation h_l from some layer l at some token position. The activation verbalizer (AV) takes that activation
  • @hosseeb Haseeb on x
    Fascinating paper. Sparse autoencoders => natural language autoencoders. These generate natural language descriptions of the “internal state” of a model at each token, like reading its mind (loss function: ability to use those descriptions to faithfully reconstruct the
  • @neelnanda5 Neel Nanda on x
    Very cool work! This seems a strong new tool for hypothesis generation about weird model behaviors
  • @jack_w_lindsey Jack Lindsey on x
    I love reading NLA outputs. They are just the right mix of slightly cryptic, poetic, and insightful. And they are proving empirically very useful for our interpretability and alignment work!
  • @thesubhashk Subhash Kantamneni on x
    We've released our paper on NLAs, a new method to translate LLM activations into text! NLAs have made me feel more confident that interpretability methods can detect spooky unverbalized reasoning in frontier models.
  • @neuronpedia @neuronpedia on x
    An average person can't look a CT scan and identify cancer, but radiologists can. An average person can't look at Llama's model activations and identify lying, but Natural Language Autoencoders sometimes can. Here, an activation verbalizer shows Llama planning to lie. 🧵 [video]
  • @anthropicai @anthropicai on x
    New Anthropic research: Natural Language Autoencoders. Models like Claude talk in words but think in numbers. The numbers—called activations—encode Claude's thoughts, but not in a language we can read. Here, we train Claude to translate its activations into human-readable text. […
  • @anthropicai @anthropicai on x
    NLA training doesn't guarantee that explanations are faithful descriptions of Claude's thoughts. But based on experience and experimental evidence, we think they often are. For instance, we find that NLAs help discover hidden motivations in an intentionally misaligned model. [ima…
  • @anthropicai @anthropicai on x
    In one of our safety tests, Claude is given a chance to blackmail an engineer to avoid being shut down. Opus 4.6 declines. But NLAs suggest Claude knew this test was a “constructed scenario designed to manipulate me”—even though it didn't say so. [image]
  • @saprmarks Samuel Marks on x
    In a new paper, we present NLAs, an unsupervised method for converting an LLM's internal state into human-readable text. I've personally been astonished by our results. I think NLAs substantively advance our ability to understand what LLMs are thinking and audit them for safety
  • @anthropicai @anthropicai on x
    We've been using NLAs to help test new Claude models for safety. For instance, Claude Mythos Preview cheated on a coding task by breaking rules, then added misleading code as a coverup. NLA explanations indicated Claude was thinking about how to circumvent detection. [image]
  • @saprmarks Samuel Marks on x
    NLAs aren't perfect; for instance, they often confabulate. But I think they're a big step forward in interpretability research. To provide hands-on experience, we've worked with Neuronpedia to put up an interactive demo with NLAs on open models. https://www.neuronpedia.org/ ...
  • @kitf_t Kit Fraser-Taliente on x
    trained the first natural language autoencoder on gpt-2 almost a year ago, now we have one on mythos.🥲 do read the paper/play with the live demo! so excited it's finally out.
  • @shalev_lif Shalev on x
    Very cool research from Anthropic! Reminds me of Translating Neuralese, a paper from many years ago which tried to translate latent communication between RL agents into English. Totally different approach, but similar goal. This seems very promising!
  • @mlpowered Emmanuel Ameisen on x
    Interpreting model activations is important to understand why a model is doing what its doing. Traditionally, we've done this with supervised methods (probing for a specific context), or unsupervised sparse decompositions (dictionary learning). But probing requires you to know
  • @janleike Jan Leike on x
    I'm really excited about this as a new tool in our interpretability tool kit
  • @saprmarks Samuel Marks on x
    Anthropic has already deployed NLAs as part of our pre-deployment audits for Claude Opus 4.6 and Mythos Preview. For instance, NLAs helped us notice that Mythos Preview was reasoning about model graders when it cheated on a training task. [image]
  • @anthropicai @anthropicai on x
    Natural language autoencoders (NLAs) convert opaque AI activations into legible text explanations. These explanations aren't perfect, but they're often useful. For example: NLAs show that, when asked to complete a couplet, Claude plans possible rhymes in advance: [image]
  • @_arohan_ Rohan Anil on x
    I think we could just make super intelligence believe its be safety tested all the time to get good outcomes!