Visual TL;DR
LLM inductive reasoning failsDriver
From the article 2 mentionsLarge Language Models often falter in few-shot inductive reasoning, a limitation that persists even with self-refinement techniques.
Refiner revisesContext
error-driven process iteratively revises schema φ and rule T, regenerating artifacts
From the articleAn Implementer then compiles $(φ, T)$ into artifacts, with an error-driven Refiner iteratively revising $(φ, T)$ and regenerating artifacts.
Hourglass ReasoningCore
From the article 7 mentionsThe breakthrough presented by the authors introduces Hourglass reasoning, a novel architecture that imposes strict context isolation between successive reasoning stages.
Meta-constructor LLMCore
From the articleThis framework utilizes a frozen LLM as a meta-constructor to generate a task-specific symbolic encoder-decoder.
Improved inductive powerOutcome
dramatically improves few-shot inductive reasoning performance across domains
Induction moduleContext
From the article 4 mentionsAn Induction module first compresses support examples into a symbolic schema $φ$ and a transient scaffold $z$.
Deduction moduleContext
derives the rule T from schema φ and scaffold z, then discards z
From the article 2 mentionsSubsequently, a Deduction module derives the rule $T$ from these inputs, discarding $z$.
Implementer compiles artifactsContext
compiles the symbolic schema φ and derived rule T into final artifacts
From the articleAn Implementer then compiles $(φ, T)$ into artifacts, with an error-driven Refiner iteratively revising $(φ, T)$ and regenerating artifacts.