Will there be a successful application of diffusion-like weight modification in LLMs before 2027? | Manifold

Will there be a successful application of diffusion-like weight modification in LLMs before 2027?

Plus

1

Ṁ1

2027

50%

chance

1D

1W

1M

ALL

This market resolves YES if a peer-reviewed research paper or major tech company publication (e.g. from Google, Meta, X.AI, OpenAI, or Anthropic) by 1/1/2027 demonstrates successful application of diffusion-like techniques (noising and trained denoising) to directly modify LLM weights usefully toward some safety or capabilities-relevant downstream purpose. Resolution will be based on papers indexed on arXiv.org or published on major AI research blogs.

This question is managed and resolved by Manifold.

#Technical AI Timelines

#Mechanistic interpretability

Get

1,000

and

3.00

Related questions

Will we have any progress on the interpretability of State Space Model LLM’s in 2024?

Will second-order optimizers displace first-order optimizers for training LLMs by 2030?

By 2028 will we be able to identify distinct submodules/algorithms within LLMs?

Will researchers extract a novel program from the weights of an LLM into a Procedural/OO programming language by 2026?

Will the best LLM in 2027 have <1 trillion parameters?

Will the best LLM in 2024 have <1 trillion parameters?

Will the best LLM in 2024 have <500 billion parameters?

Will top open-weight LLMs in 2025 reason opaquely?

Will reinforcement learning overtake LMs on math before 2028?

Will LLMs' loss function achieve the level of entropy of human text by the end of 2030?

Related questions

Will we have any progress on the interpretability of State Space Model LLM’s in 2024?

Will the best LLM in 2024 have <1 trillion parameters?

Will second-order optimizers displace first-order optimizers for training LLMs by 2030?

Will the best LLM in 2024 have <500 billion parameters?

By 2028 will we be able to identify distinct submodules/algorithms within LLMs?

Will top open-weight LLMs in 2025 reason opaquely?

Will researchers extract a novel program from the weights of an LLM into a Procedural/OO programming language by 2026?

Will reinforcement learning overtake LMs on math before 2028?

Will the best LLM in 2027 have <1 trillion parameters?

Will LLMs' loss function achieve the level of entropy of human text by the end of 2030?

© Manifold Markets, Inc.•Terms + Mana-only Terms•Privacy•Rules