REVIEW 6 cited by
The Evolution of Statistical Induction Heads: In-Context Learning Markov Chains
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
abstract
Large language models have the ability to generate text that mimics patterns in their inputs. We introduce a simple Markov Chain sequence modeling task in order to study how this in-context learning (ICL) capability emerges. In our setting, each example is sampled from a Markov chain drawn from a prior distribution over Markov chains. Transformers trained on this task form \emph{statistical induction heads} which compute accurate next-token probabilities given the bigram statistics of the context. During the course of training, models pass through multiple phases: after an initial stage in which predictions are uniform, they learn to sub-optimally predict using in-context single-token statistics (unigrams); then, there is a rapid phase transition to the correct in-context bigram solution. We conduct an empirical and theoretical investigation of this multi-phase process, showing how successful learning results from the interaction between the transformer's layers, and uncovering evidence that the presence of the simpler unigram solution may delay formation of the final bigram solution. We examine how learning is affected by varying the prior distribution over Markov chains, and consider the generalization of our in-context learning of Markov chains (ICL-MC) task to $n$-grams for $n > 2$.
Forward citations
Cited by 6 Pith papers
-
Distilled Pretraining: A modern lens of Data, In-Context Learning and Test-Time Scaling
Distilled pretraining improves test-time scaling via generation diversity but impairs induction-head-based in-context learning, with the trade-off explained by a bigram model analysis.
-
Decomposing Prediction Mechanisms for In-Context Recall
In a toy in-context recall task, label-based task initiation and observation-based continuation are distinct mechanisms with separate emergence times, and the same first-token versus second-token gap appears in an OLM...
-
Sequential Correlations Change In-Context Learning: Effective Context Length and Architectural Mismatch
Within-context token correlations reduce ICL to an effective shorter i.i.d. context length, while query–context correlations lower error and favor softmax over linear attention.
-
Unveiling the Mechanisms of Multi-Hop Reasoning in Transformers via Identity Bridge
Adding identity supervision on bridge tokens enables out-of-distribution two-hop reasoning in simple transformers, with a nuclear-norm theory explaining the benefit.
-
Selective Induction Heads: How Transformers Select Causal Structures In Context
Transformers can learn to select the correct lag of an interleaved Markov chain in context via a circuit the authors call a selective induction head, whose asymptotic optimality proof is incomplete.
-
The Features at Convergence Theorem: a first-principles alternative to the Neural Feature Ansatz for how networks learn representations
FACT is a first-order stationarity identity for weight matrices that matches or beats the Neural Feature Ansatz as a description of learned features at convergence.
Discussion (0). Continue with ORCID to comment.