Skip to content
AI News HubLIVE
Original source2 min read

Evidence Integration in Large Language Models

Summary

A new distributional theory shows that LLMs do not treat external evidence as neutral truth; instead, the model's prior beliefs shape whether and how it adopts candidate answers. Across 10 million trials and 12 models, plausible candidates are more persuasive, models absorb their own characteristic errors more readily, and identical evidence can help weaker models while hurting stronger ones—even when the model internally knows the evidence is invalid.

SourcearXiv Computational LinguisticsAuthor: Sebastien Kawada, Manolis Kellis
Evidence Integration in Large Language Models
Report an error

The correction channel is not available yet. You can copy the article reference below for later.

Correction instructions
Read article

[Submitted on 3 Sep 2026]

Title:Evidence Integration in Large Language Models

View a PDF of the paper titled Evidence Integration in Large Language Models, by Sebastien Kawada and Manolis Kellis

View PDF HTML (experimental)

Abstract:Despite increasing reliance on LLMs that reason with external evidence supplied by tools, retrieval-augmented generation, other agents, and users, how LLMs integrate such evidence into decisions they have already begun to form remains largely unclear. We present a distributional theory in which evidence shifts the receiver's distribution of initial answers, driven by a receiver prior weight and a candidate evidence tilt, leading to three predictions. First, candidates more probable to the receiver are more persuasive. Second, receivers more readily integrate characteristic errors of their own than foreign errors from different sources. Third, identical evidence can improve weaker models and harm stronger ones. We confirm these over ten million trials, twelve LLMs from four families, and eight domains, four of them scientific discovery tasks in the physical and life sciences: quantum mechanics, physics, genetics, and molecular biology. The law also yields a receiver-relative reliability frontier: receiver-congruent errors depress performance more steeply than random errors of the same rate. LLMs also integrate candidates even after internally verifying their invalidity (93-100% with propositional constraints; up to 99.4% on held-out physical and life-sciences reasoning), demonstrating evidence integration is a receiver-specific control policy over existing distributions, determined by receiver properties rather than scalar trust in the evidence source. Causal interventions show candidate integration is implemented late in the network, as a structured sequence of steps admitting external candidate answers, promoting them, and transporting them into the answer state. Representations of verification are decodable but have little causal impact on answers. A J-lens decomposition shows the state underlying verbalized verification is fully dissociable from that underlying candidate integration.

Comments: 114 pages, 16 figures, 38 tables

Subjects:

Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)

ACM classes: I.2.7; I.2.6; I.2.11; H.3.3

Cite as: arXiv:2609.04290 [cs.CL]

(or arXiv:2609.04290v1 [cs.CL] for this version)

https://doi.org/10.48550/arXiv.2609.04290

arXiv-issued DOI via DataCite

Submission history

From: Sebastien Kawada [view email] [v1] Thu, 3 Sep 2026 11:57:20 UTC (1,720 KB)

Full-text links:

Access Paper:

View a PDF of the paper titled Evidence Integration in Large Language Models, by Sebastien Kawada and Manolis Kellis

View PDF

HTML (experimental)

TeX Source

view license

Current browse context:

cs.CL

new | recent | 2026-09

Change to browse by:

cs cs.AI cs.LG

References & Citations

NASA ADS

Google Scholar

Semantic Scholar

Loading...

Data provided by:

Bibliographic Tools

Bibliographic and Citation Tools

Bibliographic Explorer Toggle

Bibliographic Explorer (What is the Explorer?)

Connected Papers Toggle

Connected Papers (What is Connected Papers?)

Litmaps Toggle

Litmaps (What is Litmaps?)

scite.ai Toggle

scite Smart Citations (What are Smart Citations?)

Code, Data, Media

Code, Data and Media Associated with this Article

alphaXiv Toggle

alphaXiv (What is alphaXiv?)

Links to Code Toggle

CatalyzeX Code Finder for Papers (What is CatalyzeX?)

DagsHub Toggle

DagsHub (What is DagsHub?)

GotitPub Toggle

Gotit.pub (What is GotitPub?)

Huggingface Toggle

Hugging Face (What is Huggingface?)

ScienceCast Toggle

ScienceCast (What is ScienceCast?)

Demos

Demos

Replicate Toggle

Replicate (What is Replicate?)

Spaces Toggle

Hugging Face Spaces (What is Spaces?)

Spaces Toggle

TXYZ.AI (What is TXYZ.AI?)

Related Papers

Recommenders and Search Tools

Link to Influence Flower

Influence Flower (What are Influence Flowers?)

Core recommender toggle

CORE Recommender (What is CORE?)

Author

Venue

Institution

Topic

About arXivLabs

arXivLabs: experimental projects with community collaborators

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.

Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?)

Key points and analysis

Article intelligence

EngineersIntermediate

Key points

  • Evidence shifts a model's prior distribution over initial answers rather than being accepted on scalar source trust alone.
  • Twelve LLMs across four families were tested: receiver-congruent errors are more readily integrated and can help weaker models but harm stronger ones.
  • Models integrated internally invalidated candidates in 93–100% of constrained settings, while verification representations had little causal effect on answers.

Highlights and analysis are generated automatically and may contain errors. Check the original source.