Will we have any progress on the interpretability of State Space Model LLM’s in 2024?
14
1kṀ2961
resolved Dec 28
Resolved
YES

State Space Models like Mamba introduce new possibilities, as States are a new object type, a compressed snapshot of a mind at a point in time which can be saved, restored, and interpreted. But a cursory search didn’t turn up any work on interpreting either States or State Space Models.

This resolves Yes if research comes out that makes any significant interpretability progress into a state space large language model. I will not bet on this market.

  • Update 2024-27-12 (PST): - Clarification:

    • The work 2404.05971v1 qualifies as making significant interpretability progress into a state space large language model. (AI summary of creator comment)

AI
LLMs
Mechanistic interpretability
Get
Ṁ1,000
to start trading!

🏅 Top traders

#NameTotal profit
1Ṁ226
2Ṁ152
3Ṁ86
4Ṁ20
5Ṁ8

People are also trading

Will LLMs be able to formally verify non-trivial programs by the end of 2025?
14% chance
By the end of 2026, will we have transparency into any useful internal pattern within a Large Language Model whose semantics would have been unfamiliar to AI and cognitive science in 2006?
10% chance
Will mechanistic interpretability be essentially solved for GPT-2 before 2030?
39% chance
Will mechanistic interpretability have more academic impact than representation engineering by the end of 2025?
72% chance
Will the state-of-the-art AI model use latent space to reason by 2026?
3% chance
Will we get a new LLM paradigm by EOY?
16% chance
Will a lab train a >=1e26 FLOP state space model before the end of 2025?
11% chance
Will interpretability be commonplace in physics papers relying on machine learning by the end of 2025?
9% chance
[Situational awareness] Will pre-2028 LLMs achieve token-output control?
38% chance
Will RL work for LLMs "spill over" to the rest of RL by 2026?
34% chance
Sort by:

Does this qualify? https://arxiv.org/abs/2404.05971v1

@CharlesFoster yes I think it should! You bet No?

bought Ṁ1,750 YES

@ahalekelly I had, because I thought it was only 50/50 that it would count. But your comment makes me think the odds are much higher, so am now betting YES.

@CharlesFoster resolved Yes, thanks for finding this!

Putting a 1k limit order for "YES" at 50% if anyone wants to take it.

I'm watching this but the notion of "any significant ... progress" is too vague for me to trade.

People are also trading

Will LLMs be able to formally verify non-trivial programs by the end of 2025?
14% chance
By the end of 2026, will we have transparency into any useful internal pattern within a Large Language Model whose semantics would have been unfamiliar to AI and cognitive science in 2006?
10% chance
Will mechanistic interpretability be essentially solved for GPT-2 before 2030?
+3% 1d39% chance
Will mechanistic interpretability have more academic impact than representation engineering by the end of 2025?
72% chance
Will the state-of-the-art AI model use latent space to reason by 2026?
3% chance
Will we get a new LLM paradigm by EOY?
16% chance
Will a lab train a >=1e26 FLOP state space model before the end of 2025?
11% chance
Will interpretability be commonplace in physics papers relying on machine learning by the end of 2025?
9% chance
[Situational awareness] Will pre-2028 LLMs achieve token-output control?
38% chance
Will RL work for LLMs "spill over" to the rest of RL by 2026?
34% chance
© Manifold Markets, Inc.•Terms•Privacy