REVIEW 4 cited by
Do LLMs dream of elephants (when told not to)? Latent concept association and associative memory in transformers
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
Signed reviews
read the original abstract
Large Language Models (LLMs) have the capacity to store and recall facts. Through experimentation with open-source models, we observe that this ability to retrieve facts can be easily manipulated by changing contexts, even without altering their factual meanings. These findings highlight that LLMs might behave like an associative memory model where certain tokens in the contexts serve as clues to retrieving facts. We mathematically explore this property by studying how transformers, the building blocks of LLMs, can complete such memory tasks. We study a simple latent concept association problem with a one-layer transformer and we show theoretically and empirically that the transformer gathers information using self-attention and uses the value matrix for associative memory.
Forward citations
Cited by 4 Pith papers
-
Stochastic Chameleons: Irrelevant Context Hallucinations Reveal Class-Based (Mis)Generalization in LLMs
LLMs systematically combine abstract category cues from a query with features from irrelevant context, causing structured answer flips, a behavior the authors call class-based (mis)generalization.
-
Associative memory inspires improvements for in-context learning using a novel attention residual stream architecture
A residual connection that copies previous-layer attention values into the current layer improves in-context learning in transformers up to 1B parameters.
-
Rethinking Associative Memory Mechanism in Induction Head
A two-layer transformer with relative positional encoding keeps its induction head active across the whole sequence, while absolute positional encoding loses it in the second half.
-
Understanding Factual Recall in Transformers via Associative Memories
A one-layer transformer can store facts at near-optimal capacity by using attention value matrices or an MLP as associative memories, and training passes through a hallucination stage.
Discussion (0). Continue with ORCID to comment.