Skip to content
OpenTrain AIFor AI Companies

HFEPX · Eval paper review

From Preference Data to Personalization: Tracing Sycophancy in Large Language Models

Himaghna Bhattacharjee, Yasha Sheynin, Shivam Singhal, Arjun Subramonian +5 more

Published

Nov 18, 2026

Citations

0

Trust level

High

Usefulness score

65/100 (Medium)

Extraction confidence

80% (High)

Derived from extracted protocol signals and abstract evidence.

Rater population

Not reported

Signals refreshed

Nov 18, 2026

Should you rely on this paper?

This paper has useful evaluation signal, but protocol completeness is partial; pair it with related papers before deciding implementation strategy.

Use this as a practical starting point for protocol research, then validate against the original paper.

Best use

Secondary protocol comparison source

Use if you need

A benchmark-and-metrics comparison anchor.

What to verify

Validate the evaluation procedure and quality controls in the full paper before operational use.

Main weakness

No major weakness surfaced.

Human feedback signal
Detected
From extracted signals
Evaluation signal
Detected
Eval setup described
Usefulness for eval research
65/100
Moderate-confidence candidate

Useful as a secondary reference; validate protocol details against neighboring papers.

Abstract

Sycophancy in large language models is often measured as agreement with the user, yet agreement can be appropriate when the user's claim is valid, their preference is genuine, or their experience warrants acknowledgment. We study sycophancy as inappropriate agreement: validating a claim, preference, or self-assessment when the context calls for correction, qualification, or resistance. We introduce a validity-aware sycophancy grader and SycoDrift, a persona-targeted evaluation set of 3,298 validity-annotated probes with controlled annotations for user openness and stakes. We use this framework to analyze three sources of agreement pressure: human preference data, automated judge specification, and in-context personalization. First, preference for validating responses in human datasets (Chatbot Arena, HH-RLHF, Community Alignment) is weak and concentrated where agreement is often appropriate. Second, in a selected-value, single-model experiment, emphasized judge rubrics produce consistent but modest increases relative to paired suppressed rubrics, demonstrating a plausible mechanism for training-time amplification. Third, in off-policy evaluations across 14 models, inference-time personalization increases inappropriate agreement, with the largest effect from matched user profiles placed directly before the query.

What we could verify

These are the protocol signals we could actually recover from the available paper metadata. Use them to decide whether this paper is worth deeper reading.

Human Feedback Types

strong

Pairwise Preference, Rubric Rating

Directly usable for protocol triage.

"Sycophancy in large language models is often measured as agreement with the user, yet agreement can be appropriate when the user's claim is valid, their preference is genuine, or their experience warrants acknowledgment."

Evaluation Modes

strong

Automatic Metrics

Includes extracted eval setup.

"Sycophancy in large language models is often measured as agreement with the user, yet agreement can be appropriate when the user's claim is valid, their preference is genuine, or their experience warrants acknowledgment."

Quality Controls

missing

Not reported

No explicit QC controls found.

"Sycophancy in large language models is often measured as agreement with the user, yet agreement can be appropriate when the user's claim is valid, their preference is genuine, or their experience warrants acknowledgment."

Benchmarks / Datasets

strong

Chatbot Arena, Anthropic HH

Useful for quick benchmark comparison.

"First, preference for validating responses in human datasets (Chatbot Arena, HH-RLHF, Community Alignment) is weak and concentrated where agreement is often appropriate."

Reported Metrics

strong

Agreement

Useful for evaluation criteria comparison.

"Sycophancy in large language models is often measured as agreement with the user, yet agreement can be appropriate when the user's claim is valid, their preference is genuine, or their experience warrants acknowledgment."

Benchmarks and datasets

Chatbot ArenaAnthropic HH

Reported metrics

agreement
Human feedback details
Uses human feedback
Yes
Feedback types
Pairwise Preference, Rubric Rating
Rater population
Not reported
Unit of annotation
Multi Dim Rubric
Expertise required
General
Evaluation details
Evaluation modes
Automatic Metrics
Agentic eval
None
Quality controls
Not reported
Evidence quality
High
Use this page as
Secondary protocol comparison source

Research brief

Metadata summary

Sycophancy in large language models is often measured as agreement with the user, yet agreement can be appropriate when the user's claim is valid, their preference is genuine, or their experience warrants acknowledgment.

Based on abstract + metadata only. Check the source paper before making high-confidence protocol decisions.

Key takeaways

  • Sycophancy in large language models is often measured as agreement with the user, yet agreement can be appropriate when the user's claim is valid, their preference is genuine, or their experience warrants acknowledgment.
  • We study sycophancy as inappropriate agreement: validating a claim, preference, or self-assessment when the context calls for correction, qualification, or resistance.
  • We introduce a validity-aware sycophancy grader and SycoDrift, a persona-targeted evaluation set of 3,298 validity-annotated probes with controlled annotations for user openness and stakes.

Researcher actions

  • Compare this paper against nearby papers in the same arXiv category before using it for protocol decisions.
  • Check the full text for explicit evaluation design choices (raters, protocol, and metrics).
  • Use related-paper links to find stronger protocol-specific references.

Caveats

  • Generated from abstract + metadata only; no PDF parsing.
  • Signals below are heuristic and may miss details reported outside the abstract.

Contribution summary

  • Sycophancy in large language models is often measured as agreement with the user, yet agreement can be appropriate when the user's claim is valid, their preference is genuine, or their experience warrants acknowledgment.
  • We study sycophancy as inappropriate agreement: validating a claim, preference, or self-assessment when the context calls for correction, qualification, or resistance.

Why it matters for eval

  • We introduce a validity-aware sycophancy grader and SycoDrift, a persona-targeted evaluation set of 3,298 validity-annotated probes with controlled annotations for user openness and stakes.

Researcher checklist

  • Human feedback protocol is explicit

    Detected: Pairwise Preference, Rubric Rating

  • Evaluation mode is explicit

    Detected: Automatic Metrics

  • Quality control reporting appears

    No calibration/adjudication/IAA control explicitly detected.

  • Benchmark or dataset anchors are present

    Detected: Chatbot Arena, Anthropic HH

  • Metric reporting is present

    Detected: agreement