# Split the Labor: Separating Evidence Interpretation from Decision Aggregation

Source: [arXiv](https://arxiv.org/abs/2608.14509v1)  
Feed7 permalink: https://feed7.dev/p/2608-14509v1-1rcfsxo  
Published: 2026-08-14T17:24:55.000Z  
Trust: Needs Review (needs_review)

## Why Included

Multi-source agents should separate source interpretation from decision aggregation. A fixed evidence tuple plus calibrated likelihood pooling avoids vote thresholds that drift as more sources arrive.

## Source Summary

The paper splits multi-source reasoning into source interpretation and decision aggregation, joined by a **four-field evidence tuple**: hypothesis, reliability bucket, rationale, and provenance. It identifies count-scale drift when unnormalized scores are summed.

## Practical Implication

For research agents, make each reader emit the same typed record, then aggregate outside the model with calibrated log-likelihood ratios. This preserves comparability, supports abstention, and prevents the decision threshold from silently moving as source count changes.

## Agent-Ready Context

The paper splits multi-source reasoning into source interpretation and decision aggregation, joined by a **four-field evidence tuple**: hypothesis, reliability bucket, rationale, and provenance. It identifies count-scale drift when unnormalized scores are summed.

For research agents, make each reader emit the same typed record, then aggregate outside the model with calibrated log-likelihood ratios. This preserves comparability, supports abstention, and prevents the decision threshold from silently moving as source count changes.

The empirical system reaches **0.921 AUPRC** versus **0.805** for a hand-crafted baseline on one longitudinal corpus. The authors also report confounded comparisons and say domain-specific quantities still require re-estimation.

## Connected Context

Feed7 judgment across 468 accumulated Signals:

This turns multi-source research into a typed interface between model judgment and deterministic aggregation. It strengthens evidence-backed approval and claim-level provenance, while explaining a specific failure of naive multi-agent voting: unnormalized scores move the effective threshold as sources accumulate. The reported gain remains confined to one corpus and requires domain recalibration.

- [Build for the Memo, Not the Demo — Shawn Chan, China Resources Holdings](https://feed7.dev/p/build-for-the-memo-not-the-demo-shawn-chan-china-resources-holdings-0i3s3oo) — The evidence tuple implements claim-level provenance, explicit reliability, and rationale in a form that can support the auditability contract proposed for finance agents.
- [Why We Killed Our Multi-Agent Pipeline — Subbiah Sethuraman and Abhilash Asokan, ZS Associates](https://feed7.dev/p/why-we-killed-our-multi-agent-pipeline-subbiah-sethuraman-and-abhilash-a-0fmz3z3) — It supports bounded specialist investigations while avoiding a fixed chain of agents: readers interpret sources independently and a separate calibrated mechanism aggregates their records.
- [What Does Done Even Mean? Agents and Paperclip's Liveness Model - Dotta, Paperclip](https://feed7.dev/p/what-does-done-even-mean-agents-and-paperclip-s-liveness-model-dotta-pap-0lx8wfc) — Calibrated aggregation and abstention provide concrete evidence for a completion or escalation decision without allowing an agent to declare success from an unsupported Boolean.
- [CausalForge: A Formally Grounded, Self-Improving Agentic Framework for Automated Research in Causal Inference](https://feed7.dev/p/2607-22511v1-0mgsdh3) — Both separate evidence processing from final judgment; CausalForge’s warning that proof does not establish intended meaning parallels the need to preserve rationale and provenance alongside aggregation.

## Context Map

- Layer: agent
- Domains: research, data
- Topics: harness-engineering, multi-agent, agent-reliability

## Uncertainty

- The empirical system reaches **0.921 AUPRC** versus **0.805** for a hand-crafted baseline on one longitudinal corpus. The authors also report confounded comparisons and say domain-specific quantities still require re-estimation.

## Agent Instruction

Use this item as source-backed context. Do not invent claims beyond the linked source. If this item conflicts with another source, call out the conflict.