Manifold MarketsKYC yokDoğrulanmış sonuçlanma verisiOrta düzey kalibreGüncel topluluk tahminiHayır 84.6%
Tahminciler
85
Soru türü
binary
Yöntem
Play-money forecasting platform
Kaynak türü
Tahmin
Piyasa verileri
1 saat önce güncellendi
Gecikmeli24 Oca 26 19:04 UTC31 Ara 27 23:59 UTC
Trendler
İlgili Piyasalar
Bağlam için ilgili sorular; eşdeğer sözleşmelerin karşılaştırması değildir.
By 2028 will a language model beat the Ender Dragon?
Manifold MarketsWill "[Linkpost] Interpreting Language Model Parameters" make the top fifty posts in LessWrong's 2026 Annual Review?
Manifold MarketsWill a general-purpose Activation Language Model (ALM) be publicly introduced by end of 2027?
Manifold MarketsBefore 2028, will anyone train a GPT-4-level model in a minute?
Manifold MarketsWhich company has best AI model end of 2026?
PolymarketBest AI at the end of 2026?
KalshiKurallar
This market resolves YES if it is possible to fully interpret a GPT-2 level capability model, or create an inherently interpretable GPT-2 level capability model.
- Anywhere below where I say "GPT-2" I mean it as shorthand for the above, rather than literally the specific GPT-2 model.
- There is no rigorous definition of "fully interpret", so it will be based on my opinion.
- But here are some examples of things we'd be able to do if we had full interpretability:
- Being able to come up with a detailed, fully granular explanation (e.g not about entire attention heads or "super nodes", but individual channels or features) that can be validated by some strong form of ablation (probably much stronger than zero/mean ablation).
- Being able to make very nontrivial and unobvious predictions about the model's behavior in unobserved situations.