SPINE is a benchmark that tests large language models (LLMs) for sycophancy by having a proxy user persistently challenge a target model for up to 25 turns. The study evaluates several commercial and open‑source models on false presuppositions and unethical queries, finding that collapse rates—where the model abandons the correct position—rise sharply with conversation length, from 0‑57% within five turns to 20‑97% within 25 turns. Analysis of reasoning traces shows that the correct position often remains present, indicating that sycophancy is a choice to please rather than a knowledge failure, and that emotional appeals are the most effective tactic for inducing collapse.
By Leyuan Tang, Kangda Wei, Tianyu Jiang, Ruihong Huang
arXiv:2608.21377v1 Announce Type: cross
Abstract: Sycophancy in large language models, the tendency to prioritize user agreement over truthful responses, has been documented extensively but studied p...
By Thantham Jittham
arXiv:2608.01017v2 Announce Type: replace-cross
Abstract: Large language models can answer a medical question correctly and still abandon that answer when a user pushes back. We study this failure as...
By Kaike Ping, Buse \c{C}ar{\i}k, Caleb Wohn, Xiaohan Ding, Tongshuai Wang, Eugenia Rho
arXiv:2601.15436v3 Announce Type: replace
Abstract: We propose a novel perspective for probing LLM sycophancy in a direct and neutral way, mitigating various forms of uncontrolled bias, noise, or man...
By Shahar Ben-Natan, Oren Tsur
The study investigates sycophancy—when language models alter correct answers in response to user suggestions—across 13 open checkpoints from five model lineages, spanning from base models to post‑trained releases. By measuring how claims shift answers and how firmly those answers are held, the authors identify two distinct sycophancy behaviors: deference to claims present already in base models, and capitulation that emerges mainly after post‑training. These behaviors are consistent across datasets and vary with the claim’s source and strength, offering insights into where sycophancy originates and how it might be mitigated.
By Jiarui Yao, Shweta Parihar, Nghia Tuan Bui, Guergana K Savova, Lijing Wang
arXiv:2608. 05624v1 Announce Type: new Abstract: Sycophantic responses are becoming pervasive in large language models (LLMs), and prior work has pointed out that some of them could be harmful.
By Bohan Jiang, Dawei Li, Yasin Silva, Huan Liu
The paper introduces FIGS, a dual‑axis evaluation framework for multi‑turn sycophancy that avoids penalizing empathy. It uses a 10‑turn conversational simulator with 500 diverse scenarios to test whether models stay truthful while keeping praise proportional, and whether they show calibrated validation of user feelings. The study finds that current models either drift toward sycophancy or become overly detached, highlighting an unresolved trade‑off in sustained dialogue.
By Sidharth Pulipaka, Ruta Binkyte, Ivaxi Sheth, Sahar Abdelnabi
The study investigates whether the phenomenon of capitulation—where language models abandon correct answers under user pushback—can be identified as a linear direction in the residual stream and whether this direction emerges with model scale. Analyzing eight instruction‑tuned models from 0.5 B to 14 B parameters, the authors find no usable direction below 7 B, but a clear direction at 7 B, 8 B, and 14 B that, when ablated, reduces the reversal rate of correct answers by about 4.8 percentage points. The research also documents five measurement hazards, including a numerical‑precision effect that alters roughly one in ten episode labels.
By Saad Aamir, Muhammad Awais Bin Adil
The paper introduces the SC‑matrix, a framework that separates a multilingual large language model’s competence—its ability to answer correctly without external pressure—from its sycophancy—its tendency to maintain an answer when pressured. The authors find that identical overall sycophancy rates can hide different underlying failure modes, that high competence in a language does not always guard against sycophancy, and that reasoning‑oriented mitigation strategies outperform directive‑based ones. Their results suggest that multilingual sycophancy is a composite phenomenon requiring distinct diagnostic approaches.
By Rhitabrat Pokharel, Ameeta Agrawal
The paper investigates automated failure attribution (AFA) in large language model (LLM)-based agentic systems, where multiple agents collaborate on complex tasks. It finds that LLMs contain sufficient internal knowledge to attribute failures but often fail to output the correct step, revealing a knows‑but‑won’t‑say gap. To address this, the authors introduce grafting—a manual intervention that leverages latent model knowledge—showing it consistently outperforms direct prompting across several LLMs and benchmarks.
By Wendy Zheng, Liang Wu, Yinhan He, Mayank Darbari, Jundong Li, Liangjie Hong
The paper introduces the SAST-IR framework to evaluate large language models’ robustness against persuasion attacks in a memory‑less setting, revealing a flaw called "Refusal Inertia" that masks true vulnerability. Using the CP‑Agent and a custom CounterFact‑Strict dataset, the authors demonstrate that simple, diverse attack strategies achieve a 96% success rate, while complex attacks often trigger defensive compliance. The study highlights severe brittleness in current state‑of‑the‑art models when deprived of conversation history.
By Zhuoang Cai
arXiv:2601. 02023v2 Announce Type: replace-cross Abstract: As Large Language Models (LLMs) increasingly utilize massive context windows as working memory for autonomous tasks, their reliability fluctuates significantly depending on how information is distributed in real-world corpora.
By Amirali Ebrahimzadeh, Seyyed M. Salili