Inference cost

Computational expense proportional to number of generated tokens, targeted for reduction by ReflCtrl

Neighborhood — ranked by edge-count

claim

Reflections are redundant in many cases, especially in stronger models
associated_with
Key interpretive finding that stronger models can have reflections reduced with minimal accuracy cost

concept

Self-reflection
associated_with
The ability of reasoning LLMs to review and revise previous reasoning steps during inference

cosine ≥ 0.65 · no typed edge

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

Expected Costconcept0.780
The KL divergence between predicted and preferred outcomes, minimized by policies that realize prior preferences.
Inference-Time Scalingconcept0.770
Paradigm of improving model outputs through more computation at inference time; VS is presented as a diversity-oriented alternative
Stages of Inferenceframework0.770
The perspective that LLM inference decomposes into distinct computational stages, which the paper extends to looped models
perceptual inferenceconcept0.757
The process of inferring causes of sensory inputs, a key aspect of the free-energy minimization scheme.
Bayesian Inferencemethod0.756
Inference to the Best Explanationconcept0.749
Methodological principle applied to favor identity over correlation between signed evaluation and felt valence
inference of sentienceconcept0.746
Attributing subjective experience based on observable embodied behaviours.
Embodied Inferenceconcept0.743
Process by which organism's material states and internal dynamics realize variational inference through action