CARTBench is a new benchmark for evaluating large language model shopping assistants on complex, ambiguous requests that involve timelines. It compiles 1,603 timestamped dialogues from 17 retail domains and pairs them with product records, then uses case‑specific rubrics based on Ideal Response Principles to score responses across Core, Excellence, and Safety criteria. In a full‑suite evaluation of ten commercial APIs and several open‑weight models, only three models scored above 90% on Core and none surpassed 45% on Excellence, revealing a gap between required behavior and high‑quality purchase advice.
arXiv:2606. 17698v1 Announce Type: new Abstract: As LLM-based shopping agents enter production, existing benchmarks fail to capture how a shopper's requirements arrive: stated implicitly in the query, recorded in a profile, or revealed only when the right question is asked.
RealWorldShop introduces a new benchmark for conversational shopping agents, featuring 3.28 million grounded products, structured shopping episodes, a profile‑grounded user simulator, and role‑play evaluation. Analysis reveals that existing systems generate locally plausible responses but struggle with state tracking, constraint updating, and grounded convergence, especially in ambiguous or multi‑intent scenarios. The authors propose REALSHOP_AGENT, a session‑control framework with explicit state management, shopping‑flow control, catalog‑grounded retrieval, and runtime guards, which consistently outperforms strong baselines on the benchmark.
arXiv:2606. 10156v1 Announce Type: cross Abstract: As recommender systems transition toward agentic, multi-turn conversational interfaces, evaluation paradigms have struggled to keep pace.
PACEShop introduces a new evaluation framework, PACE, for shopping assistants that emphasizes personalized, actionable, compositional, and evidence‑grounded responses. The benchmark dataset contains 22,625 records with structured personas, auditable evidence pools, and detailed defect annotations, while PACEJudge offers a training‑free protocol for assessing these dimensions. Experiments demonstrate that generic judges miss key diagnostic fields, whereas PACEJudge improves evaluation across persona alignment, cross‑component consistency, grounding, and defect localization without retraining.
The paper presents a method for fine‑tuning a large language model (LLM) recommender to generate personalized, non‑harmful explanations for its recommendations. By training two LLM‑judge reward models and using constrained GRPO, the authors achieve a significant increase in the PASS rate for all three criteria, from 0.649 to 0.956 on their own judges and from 0.677 to 0.931 on an independent judge. The fine‑tuned model maintains its original recommendation performance, demonstrating that LLM‑based recommenders can be adapted to complex tasks without loss of effectiveness.
MOSS (Model Sanity Service) is a production framework that embeds LLM-based relevance scoring into a continuous monitoring loop for recommendation, search, and offer surfaces, using targeted sampling, strategy-aware prompting, and structured alerting to diagnose semantic failures. It has been deployed for over six months at a large U.S. retailer, identifying two confirmed serving defects and quantifying a 76.4% reduction in alert rate for a backfill blending fix. Benchmark results show that LLM judges outperform non-LLM baselines on both recommendation and search tasks, demonstrating that a single LLM monitoring framework can deliver actionable quality signals across multiple product discovery surfaces.
The paper introduces a deterministic, reproducible e‑commerce environment that pre‑commits customer and trajectory parameters, enabling a simulated consumer to attempt purchasing a target cart with the help of an evaluated model. The environment records every assistant action and state, allowing post‑trial evaluation of specific conversation components and applying penalties based on tool‑call accuracy. Using this setup, the authors benchmark eight open‑weight agents (20B–35B parameters) across 160 trials and 44 metrics, revealing nuanced performance issues such as under‑action, over‑purchase, unsupported product attributes, and poor search that are hidden by overall success rates.
arXiv:2308.12241v2 Announce Type: replace-cross Abstract: Recently, the fast development of Large Language Models (LLMs) such as ChatGPT has significantly advanced NLP tasks by enhancing the capabili...
arXiv:2606. 12924v1 Announce Type: new Abstract: We present a modular two-agent simulation framework for evaluating conversational shopping assistant architectures.
arXiv:2608. 14068v1 Announce Type: cross Abstract: Conversational recommendation for e-commerce is increasingly mediated by large language models (LLMs), yet many real-world deployments operate under a stricter requirement: recommendations must be drawn only from a merchant's fixed catalog, without web search or unsupported product claims.
The paper introduces SalesLLM, a bilingual (Chinese/English) benchmark for evaluating large language models (LLMs) in realistic sales dialogues. It comprises 30,074 scripted configurations and 1,805 curated multi‑turn scenarios from Financial Services and Consumer Goods, with controllable difficulty and personas. An automatic evaluation pipeline uses an LLM judge for sales‑process progress and fine‑tuned BERT classifiers for end‑of‑dialogue buying intent, while a user model, CustomerLM, is trained to improve simulation fidelity. SalesLLM scores correlate strongly with human ratings (Pearson r = 0.86) and reveal that top Chinese LLMs match junior‑to‑intermediate human salespeople but not experts, with cross‑lingual consistency remaining poor.
RPCBench is a new benchmark designed to evaluate large language models’ ability to critique recommendation requests by detecting, diagnosing, and handling flawed premises. It includes evidence‑grounded test instances across five recommendation domains and ten types of premise failures, and introduces a fine‑grained evaluation framework covering detection, error localization, handling strategy, and evidence faithfulness. Experiments with 11 LLMs reveal that proactive detection is the main bottleneck, with models struggling most on underspecified‑premise errors and showing that optimal critique quality occurs at intermediate reasoning lengths.