arXiv AI

When Both Sides Adapt: The Challenges of Evaluating Interactive LLM Systems

The paper investigates how user adaptivity in interactive LLM systems affects evaluation outcomes. It distinguishes between the common‑distribution effect (model performance on a fixed prompt set) and the induced‑distribution effect (performance on prompts generated by user interaction). The authors show that offline benchmarks and live A/B tests target different estimands, can rank models differently, and that user adaptivity can increase estimation variance and lead to challenges such as reward hacking and novelty effects.

arXiv AI
Oct 15

Interaction-as-a-Judge

arXiv:2610.17895v2 Announce Type: cross Abstract: Coding agents are tasked with producing sophisticated interactive applications such as websites, desktop applications, and games. While humans assess...

By Wayne Chi, I-Hung Hsu, Li Chen, Wei-Lin Chiang, Ion Stoica, Anastasios Nikolas Angelopoulos, Ameet Talwalkar, Chris Donahue, Cho-Jui Hsieh
arXiv AI
Jul 24

Benchmarking the Personalization Capabilities of Large Language Models

arXiv:2607. 20471v1 Announce Type: new Abstract: Personalization, the act of varying a message to induce action from a specific receiver while keeping sender, channel, and time fixed, has a long tradition in psychology and marketing as a two-party problem in which sender and receiver have independent objectives.

By Ashutosh Srivastava, Siddharth Yedlapati, Vinay Aggarwal, Yaman Kumar Singla, Shashwat Dixit, Jitendra Ajmera, Balaji Krishnamurthy
arXiv Machine Learning
Sep 1

Adaptive Doubly Robust Off-Policy Evaluation for Ranking Policies under Diverse User Behavior

The paper introduces Adaptive Doubly Robust (ADR), an off‑policy evaluation method for ranking policies that blends adaptive importance weighting with reward regression to reduce variance. ADR is unbiased when the true user behavior model is known and, under a sufficient condition, achieves lower variance than the prior Adaptive Inverse Propensity Scoring (AIPS) approach. Experiments on synthetic data show that ADR consistently improves mean squared error over AIPS and other ranking OPE estimators across various data sizes and ranking lengths.

By Kosuke Iguchi, Ren Kishimoto
arXiv Computation and Language
Oct 9

UXBench Pro: Benchmarking Personalized User Experience in Multi-Turn Dialogue Interactions

UXBench Pro is a new benchmark that evaluates personalized user experience in multi-turn dialogue interactions. It contains 1,000 test instances from real user interactions across 12 task scenarios and 82 domains, each paired with a FACTORS user profile describing seven behavioral facets. The benchmark introduces a dual-perspective evaluation: a personalized User Reward Model for third‑person judgment and Sim4Eval, a user simulator for first‑person evaluation across four cognitive state dimensions, along with meta‑benchmarks URMBench and USimBench to assess evaluator reliability.

By Mengze Hong, Zeyang Lei, Wenbo Shang, Xia Zeng, Xiying Zhao, Qi Zhu, Chen Jason Zhang, Di Jiang, Taiming Fu, Qiongyi Zhou, Qinghe Chang, Fubao Zhang, Chenxuan Ma, Minlong Peng, Jinfeng Huang, Zineng Zhou, Jindou Wu, Muge Qi, Sijun He, Xin Cui, Di Liang, Yuan Hua, Davey Chen
arXiv Machine Learning
Jul 31

Personalized RewardBench: Evaluating Reward Models with Human Aligned Personalization

arXiv:2604. 07343v2 Announce Type: replace-cross Abstract: Pluralistic alignment has emerged as a critical frontier in the development of Large Language Models (LLMs), with reward models (RMs) serving as a central mechanism for capturing diverse human values.

By Qiyao Ma, Dechen Gao, Rui Cai, Boqi Zhao, Hanchu Zhou, Junshan Zhang, Zhe Zhao
arXiv Computation and Language
Oct 23

What Pairwise Evaluation Misses: Reward Scores across Prompts

The paper investigates how reward models, used for preference optimization and ranking, can be biased by prompt-specific effects that are not captured by pairwise evaluation. By comparing between‑prompt variance in model scores to human ratings, the authors find that reward scores cluster strongly by prompt, often more so than human judgments. They propose a human‑referenced audit that examines between‑prompt variation, sensitivity to incorrect prompts, and usefulness for response selection, and demonstrate that adjusting for prompt means can reduce misleading global selection of responses.

By Mateusz Modrzejewski, Tomasz Radzikowski