Skip to content

Sumcheck optimization: eq-elimination for base-field claims #170

@psivesely

Description

@psivesely

When a sumcheck polynomial has the form $g(\mathbf{x}) = \widetilde{\mathsf{eq}}(\mathbf{r}, \mathbf{x}) \cdot p(\mathbf{x})$, the prover can exploit the tensor structure of the eq factor. Since $\widetilde{\mathsf{eq}}(\mathbf{r}, \mathbf{x}) = \prod_i (r_i x_i + (1 - r_i)(1 - x_i))$ factors across coordinates, the prover can maintain a running eq-table and fold in one scalar factor per round rather than treating eq as a general bookkeeping table. This "nearly eliminates" the prover cost of the eq factor. See Section 5 of Speeding Up Sum-Check Proving (Extended Version).

More generally, consider a sumcheck over $g(\mathbf{x}) = \sum_i p_i(\mathbf{x}) q_i(\mathbf{x})$ where only some $q_i$ are eq polynomials. The prover can split the sum into an eq-structured group and an unstructured group, apply eq-elimination to the first group, and handle the second group with standard bookkeeping. This pattern arises naturally for example in the incoming batched tensor evaluation algorithm.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions