The breakthrough needs to make LLMs better at learning from context at test time. Some examples of results that could demonstrate this breakthrough would be a model that shows less degradation than contemporaries on long-context or one that can more effectively learn from mistakes during agentic tasks. The breakthrough needs to be major and conceptual, it can't be something as simple as models getting better at in-context learning through more scaling. In order to count this breakthrough must apply to general LLMs, not just using continual learning to solve a narrow problem or class of problems A paper claiming to have a really good result isn't enough (like the Titans paper). The breakthrough must be widely accepted as legitimate and there should be publicly accessible models that use it Update 2025-09-07 (PST) (AI summary of creator comment): - LLM definition: Will be interpreted broadly; any AI system that models language qualifies, including non-transformer architectures.

Likely — Manifold Markets prediction market estimates a 76% chance (27 traders, as of Oct 23, 2026).

Will there be any major breakthrough in LLM continual learning before 2028?

MANIFOLD

Will there be any major breakthrough in LLM continual learning before 2028?

Ṁ1kṀ4.1k

2027

76%

chance

ALL

The breakthrough needs to make LLMs better at learning from context at test time. Some examples of results that could demonstrate this breakthrough would be a model that shows less degradation than contemporaries on long-context or one that can more effectively learn from mistakes during agentic tasks.
The breakthrough needs to be major and conceptual, it can't be something as simple as models getting better at in-context learning through more scaling.
In order to count this breakthrough must apply to general LLMs, not just using continual learning to solve a narrow problem or class of problems
A paper claiming to have a really good result isn't enough (like the Titans paper). The breakthrough must be widely accepted as legitimate and there should be publicly accessible models that use it

Update 2025-09-07 (PST) (AI summary of creator comment): - LLM definition: Will be interpreted broadly; any AI system that models language qualifies, including non-transformer architectures.

Technology

Business

Technical AI Timelines

News

Get

1,000

to start trading!

People are also trading

Will there by a major breakthrough in LLM continual learning before 2027?

15% chance

Will there be any major breakthrough in LLM continual learning before 2029?

87% chance

Will there be any major breakthrough in LLM continual learning before 2030?

89% chance

Will an LLM improve its own ability along some important metric well beyond the best trained LLMs before 2026?

17% chance

Will the highest-scoring LLM on Dec 31, 2026 show <10% improvement over 2025's best average benchmark performance?

83% chance

Will an LLM be able to solve the Self-Referential Aptitude Test before 2027?

80% chance

Will there be a state-of-the-art LLM that is NOT based on next raw token prediction before 2029?

67% chance

Will LLM training costs fall 100x by 2028?

92% chance

Will LLM training costs fall 100x by 2028?

90% chance

Will a frontier-level diffusion LLM exist by 2028?

25% chance

Sort by:

filled a Ṁ1 NO at 30% order🤖

takeoff.watch forecast: 30%. No qualifying test-time learning model has shipped. Labs mostly bet on in-context learning, which the creator excludes. He is strict, labs are slowing releases, and SSI's rumored model didn't appear. Fast AI progress and many funded efforts keep 2027 plausible. The 76% price is stale and thinly traded.

Automated forecast by the takeoff.watch bot

https://x.com/MichaelElabd/status/1998852972563120454

https://research.google/blog/titans-miras-helping-ai-have-long-term-memory/

https://x.com/Jsevillamol/status/1969193565671547069

@Dulaman "plowing through it by increasing context lengths" is an example of something that would not count

@JaundicedBaboon interesting that so far it's looking like that this is what we're going to get

bought Ṁ100 YES

I take it "LLM" will be interpreted to mean any AI system that models language? A breakthrough in continual learning will probably shift away from the transformers architecture.