Claude
Channel @anthropicai
Anthropic is an AI safety and research company that’s working to build reliable, interpretable, steerable AI systems, and the creators of the Claude Chat Bot Assistant.
Rating out of 10 4.6 Rating out of 10 4.6 /10 System score · How ratings work
- Subscribers
- 806
- Last post
- 1y ago
Collected on 2026-11-21 · not real-time
Introduction
Anthropic is an AI safety and research company that’s working to build reliable, interpretable, steerable AI systems, and the creators of the Claude Chat Bot Assistant.
Show the original Telegram description
Anthropic is an AI safety and research company that’s working to build reliable, interpretable, steerable AI systems, and the creators of the Claude Chat Bot Assistant.
Rating breakdown
Algorithm v1 · How ratings workSystem score 4.6 (1 + contributions 3.56) · final 4.6
Dimension Input Value Weight Contribution
Content quality Not reviewed yet (counted as neutral 0.5) · 40% 0.50 40% +1.80
Safety Not reviewed yet (counted as neutral 0.5) · 30% 0.50 30% +1.35
Audience 806 subscribers · 12% 0.30 12% +0.32
Recent activity Last post 1y ago · 10% 0.10 10% +0.09
Telegram verified Not verified · 4% 0.00 4% —
Official backlink No linked tools · 4% 0.00 4% —
Recent posts
View all on TelegramChannel name was changed to «Claude»
[Media]
10.9K views
Models struggle to transition between these strategies, as exhibited by a spike in test loss. This spike moves to larger datasets as one increases model capacity. This is a clear signature of double-descent, a phenomenon that is now well-known in the ML literature.
We hope these results are a step towards a mechanistic theory of memorization. There are many open questions, such as understanding the loss spike, or what happens when only a subset of the data is repeated.
Thanks to Adam Jermyn for his comments reproducing and extending these results!
https://transformer-circuits.pub/2023/toy-d…
9.8K views Media
For small training sets, models use superposition to memorize more data points than the two available neurons. For large training sets, models learn features in superposition, as observed in our previous work, allowing the model to generalize.
https://transformer-circuits.pub/2022/toy_model/index.html
6.9K views Media
Our prior work showed that these toy models use a strategy called “superposition” to learn more features than available neurons. Here we observe how training data points, as well as features, are embedded in the hidden space.
5.7K views Media
Similar Telegram resources
Spotted an error? Suggest an update
About the data
Counts, verification and recent posts come from periodic snapshots of Telegram's public pages and are not real-time. Descriptions are edited; the original text is available under the original description.