arXiv subjects

Fernando Diaz

Publications and source records attributed to Fernando Diaz.

At least 19 recordsLinked to original sources

arXiv2026

"I want to be pushed, I want to grow": Enabling social workers to design evaluations of LLM augmentation in their work

Workers are increasingly asked to adopt AI systems to assist their work, yet are rarely given a voice in defining what meaningful AI augmentation should look like or how to evaluate for it. In this paper, we propose worker-driven AI measurement---a bottom-up approach to AI evaluation where workers collaboratively shape decisions about which tasks AI should augment, what "successful" augmentation looks like, and how it should be measured. We explore how to support this through a case study with 19 workers from a local school social work organization. Through a series of eight workshops, workers iteratively develop their own measurement goals for AI evaluation, systematize these goals, and then design a benchmark to capture how effectively an LLM can "challenge" them to reflect on their own assumptions and biases in the context of their day-to-day work. Workers collaboratively design and refine an LLM-as-a-judge rubric based on their professional and lived expertise. In validations of the worker-created benchmark, we find that there is strong agreement between worker and LLM judge ratings and that the resulting benchmark can differentiate performance across six state-of-the-art LLMs. Based on our case study, we discuss opportunities for future work to support worker-driven AI measurement as a complementary approach to existing top-down AI evaluation approaches.

cs.HC↗

arXiv2026

Multi-Agent Transactive Memory

The decentralized deployment of LLM agents with diverse capabilities across diverse tasks motivates infrastructure for knowledge sharing across heterogeneous agent populations. Just as search engines index human-generated artifacts to support human problem solving, retrieval systems can organize agent-generated artifacts for reuse across agent populations. We extend retrieval-augmented generation - which demonstrates the value of human-authored artifacts to individual agents - to retrieval of agent-generated artifacts supporting a population of agents. In particular, agent trajectories encode reusable procedural knowledge, yet these artifacts are typically discarded after a single use or retained only by the producing agent, forcing newly instantiated agents to repeatedly rediscover existing solutions. We propose Multi-Agent Transactive Memory (MATM), a framework for population-level storage and retrieval of agent-generated trajectories, where producer agents contribute trajectories to a shared repository and consumer agents retrieve them to improve task execution. We focus on interactive environments (ALFWorld and WebArena), where trajectories are long and encode especially rich procedural structure. Our experiments demonstrate that retrieving trajectories from MATM improves downstream task performance and reduces interaction steps without coordination or joint training. These results position MATM as a design pattern for population-level experience sharing in open agent ecosystems.

cs.AI↗

arXiv2026

Offline Preference-Based Trajectory Evaluation

Offline evaluation of agentic systems often collapses trajectories to terminal success, discarding information about partial progress and inducing widespread ties, creating substantial statistical inefficiency by reducing effective sample size and weakening the ability to distinguish systems. We propose preference-based trajectory evaluation, which compares trajectories directly through temporal preferences over progress and time-to-return profiles. We find that, across diverse agentic and interactive benchmarks, standard success-based metrics produce tied comparisons on roughly 75% of instances, whereas trajectory-aware preferences reduce ties to roughly 35%, improving discriminative power, ranking stability, and data efficiency. Our results suggest that benchmark saturation, often attributed to poor data collection or problem difficulty, may also be explained by the choice of evaluation measure.

cs.LG↗

arXiv2026

Characterizing Cultural Localization in AI-Generated Stories

The global use of artificial intelligence has increased interest in assessing the ability to generate culturally localized content, including stories. Cultural localization in stories often occurs through either templated localization -- the use of cultural markers (e.g., names, locations) in a generic narrative -- or holistic localization -- the variation of plots, values, and themes, in addition to cultural markers. We propose a method to measure the degree to which content was generated through templated localization. Specifically, we identify the lexical tokens that distinguish stories across nationalities and measure the similarity of the narratives that remain after removing them. In stories generated by five models on 125 topics for 193 nationalities, our method is able to detect that only a small subset (9-17%) of the vocabulary accounts for the variation across nationalities and that the narratives that remain after removing them contain repeated multi-word sequences, suggesting the presence of a shared culturally-agnostic narrative template. Finally, we characterize the cultural markers for their stereotypicality and offensiveness, finding that markers from 19 countries, mostly located in the Global South, are on average offensive.

cs.CL↗

arXiv2026

GrepSeek: Training Search Agents for Direct Corpus Interaction

Large Language Model (LLM) search agents have shown strong promise for knowledge-intensive language tasks through multiple rounds of reasoning and information retrieval. Most existing systems access information using a retriever that takes a keyword or natural language query and returns a ranked list of documents using an index of pre-computed document representations. In this work, we explore a complementary perspective in which the search agent treats the corpus itself as the search environment and finds evidence by issuing executable shell commands. We introduce GrepSeek, an optimized direct corpus interaction (DCI) search agent that trains a compact search agent to find, filter, and compose evidence from large text corpora. To address the instability of learning behavior directly with reinforcement learning on large corpora, we propose a two-stage training pipeline. First, we construct a cold-start dataset using an answer-aware Tutor and answer-blind Planner to generate verified, causally grounded search trajectories. Second, we refine the initialized policy with Group Relative Policy Optimization (GRPO), allowing the agent to improve its task-oriented search behavior through direct interaction with the corpus. To make DCI practical at scale, we further use a semantics-preserving sharded-parallel execution engine that accelerates shell-based retrieval by up to $7.6\times$ while preserving byte-exact equivalence with sequential execution of the shell command. Experiments across seven open-domain question answering benchmarks show that GrepSeek achieves the strongest overall token-level $F_1$ and Exact Match. Our analysis also highlights the limitations of purely lexical interaction on queries with substantial surface-form variation, suggesting DCI as a practical and competitive method for search agents that can complement existing retrieval paradigms in the real world.

cs.CL↗

arXiv2026

Quantifying the Statistical Effect of Rubric Modifications on Human-Autorater Agreement

Autoraters, also referred to as LLM-as-judges, are increasingly used for evaluation and automated content moderation. However, there is limited statistical analysis of how modifications in a rubric presented to both humans and autoraters affect their score agreement. Rubrics that ask for an overall or \emph{holistic} judgment - for example, rating the ``quality'' of an essay - may be inconsistently interpreted due to the complexity or subjectivity of the criteria. Conversely, rubrics can ask for \emph{analytic} judgments, which decompose assessment criteria - for example, ``quality'' into ``fluency'' and ``organization''. While these rubrics can be edited to improve the individual accuracy of both human and automated scoring, this approach may result in disagreement between the two scores, or with the associated holistic judgment. Designing and deploying reliable autoraters requires understanding not just the relationship between human and autorater annotations but how that relationship changes as holistic or analytic judgments are elicited. The results indicate that rubric edits providing representative examples and additional context, and reducing positional bias in the rubric increased human-autorater agreement, while higher rubric complexity and conservative aggregation methods tended to decrease it. The findings from the automatic essay scoring and instruction-following evaluation domains suggest that practitioners should carefully analyze domain- and rubric-specific performance to move towards higher human-autorater agreement.

cs.CL↗

arXiv2026

Multilingual and Domain-Agnostic Tip-of-the-Tongue Query Generation for Simulated Evaluation

Tip-of-the-Tongue (ToT) retrieval benchmarks have largely focused on English, limiting their applicability to multilingual information access. In this work, we construct multilingual ToT test collections for Chinese, Japanese, Korean, and English, using an LLM-based query simulation framework. We systematically study how prompt language and source document language affect the fidelity of simulated ToT queries, validating synthetic queries through system rank correlation against real user queries. Our results show that effective ToT simulation requires language-aware design choices: non-English language sources are generally important, while English Wikipedia can be beneficial when non-English sources provide insufficient information for query generation. Based on these findings, we release four ToT test collections with 5,000 queries per language across multiple domains. This work provides the first large-scale multilingual ToT benchmark and offers practical guidance for constructing realistic ToT datasets beyond English.

cs.IR↗

arXiv2026

Evaluation of Agents under Simulated AI Marketplace Dynamics

Modern information access ecosystems consist of mixtures of systems, such as retrieval systems and large language models, and increasingly rely on marketplaces to mediate access to models, tools, and data, making competition between systems inherent to deployment. In such settings, outcomes are shaped not only by benchmark quality but also by competitive pressure, including user switching, routing decisions, and operational constraints. Yet evaluation is still largely conducted on static benchmarks with accuracy-focused measures that assume systems operate in isolation. This mismatch makes it difficult to predict post-deployment success and obscures competitive effects such as early-adoption advantages and market dominance. We introduce Marketplace Evaluation, a simulation-based paradigm that evaluates information access systems as participants in a competitive marketplace. By simulating repeated interactions and evolving user and agent preferences, the framework enables longitudinal evaluation and marketplace-level metrics, such as retention and market share, that complement and can extend beyond traditional accuracy-based metrics. We formalize the framework and outline a research agenda, motivated by business and economics, around marketplace simulation, metrics, optimization, and adoption in evaluation campaigns like TREC.

cs.IR↗

arXiv2026

Overview of the TREC 2025 Tip-of-the-Tongue track

Tip-of-the-tongue (ToT) known-item retrieval involves re-finding an item for which the searcher does not reliably recall an identifier. ToT information requests (or queries) are verbose and tend to include several complex phenomena, making them especially difficult for existing information retrieval systems. The TREC 2025 ToT track focused on a single ad-hoc retrieval task. This year, we extended the track to general domain and incorporated different sets of test queries from diverse sources, namely from the MS-ToT dataset, manual topic development, and LLM-based synthetic query generation. This year, 9 groups (including the track coordinators) submitted 32 runs.

cs.IR↗

arXiv2025

Diversification as Risk Minimization

Users tend to remember failures of a search session more than its many successes. This observation has led to work on search robustness, where systems are penalized if they perform very poorly on some queries. However, this principle of robustness has been overlooked within a single query. An ambiguous or underspecified query (e.g., ``jaguar'') can have several user intents, where popular intents often dominate the ranking, leaving users with minority intents unsatisfied. Although the diversification literature has long recognized this issue, existing metrics only model the average relevance across intents and provide no robustness guarantees. More surprisingly, we show theoretically and empirically that many well-known diversification algorithms are no more robust than a naive, non-diversified algorithm. To address this critical gap, we propose to frame diversification as a risk-minimization problem. We introduce VRisk, which measures the expected risk faced by the least-served fraction of intents in a query. Optimizing VRisk produces a robust ranking, reducing the likelihood of poor user experiences. We then propose VRisker, a fast greedy re-ranker with provable approximation guarantees. Finally, experiments on NTCIR INTENT-2, TREC Web 2012, and MovieLens show the vulnerability of existing methods. VRisker reduces worst-case intent failures by up to 33% with a minimal 2% drop in average performance.

cs.IR↗

arXiv2025

Taxonomy of User Needs and Actions

The growing ubiquity of conversational AI highlights the need for frameworks that capture not only users' instrumental goals but also the situated, adaptive, and social practices through which they achieve them. Existing taxonomies of conversational behavior either overgeneralize, remain domain-specific, or reduce interactions to narrow dialogue functions. To address this gap, we introduce the Taxonomy of User Needs and Actions (TUNA), an empirically grounded framework developed through iterative qualitative analysis of 1193 human-AI conversations, supplemented by theoretical review and validation across diverse contexts. TUNA organizes user actions into a three-level hierarchy encompassing behaviors associated with information seeking, synthesis, procedural guidance, content creation, social interaction, and meta-conversation. By centering user agency and appropriation practices, TUNA enables multi-scale evaluation, supports policy harmonization across products, and provides a backbone for layering domain-specific taxonomies. This work contributes a systematic vocabulary for describing AI use, advancing both scholarly understanding and practical design of safer, more responsive, and more accountable conversational systems.

cs.HC↗

arXiv2025

RankList -- A Listwise Preference Learning Framework for Predicting Subjective Preferences

Preference learning has gained significant attention in tasks involving subjective human judgments, such as \emph{speech emotion recognition} (SER) and image aesthetic assessment. While pairwise frameworks such as RankNet offer robust modeling of relative preferences, they are inherently limited to local comparisons and struggle to capture global ranking consistency. To address these limitations, we propose RankList, a novel listwise preference learning framework that generalizes RankNet to structured list-level supervision. Our formulation explicitly models local and non-local ranking constraints within a probabilistic framework. The paper introduces a log-sum-exp approximation to improve training efficiency. We further extend RankList with skip-wise comparisons, enabling progressive exposure to complex list structures and enhancing global ranking fidelity. Extensive experiments demonstrate the superiority of our method across diverse modalities. On benchmark SER datasets (MSP-Podcast, IEMOCAP, BIIC Podcast), RankList achieves consistent improvements in Kendall's Tau and ranking accuracy compared to standard listwise baselines. We also validate our approach on aesthetic image ranking using the Artistic Image Aesthetics dataset, highlighting its broad applicability. Through ablation and cross-domain studies, we show that RankList not only improves in-domain ranking but also generalizes better across datasets. Our framework offers a unified, extensible approach for modeling ordered preferences in subjective learning scenarios.

cs.LG↗

arXiv2025

MoR: Better Handling Diverse Queries with a Mixture of Sparse, Dense, and Human Retrievers

Retrieval-augmented Generation (RAG) is powerful, but its effectiveness hinges on which retrievers we use and how. Different retrievers offer distinct, often complementary signals: BM25 captures lexical matches; dense retrievers, semantic similarity. Yet in practice, we typically fix a single retriever based on heuristics, which fails to generalize across diverse information needs. Can we dynamically select and integrate multiple retrievers for each individual query, without the need for manual selection? In our work, we validate this intuition with quantitative analysis and introduce mixture of retrievers: a zero-shot, weighted combination of heterogeneous retrievers. Extensive experiments show that such mixtures are effective and efficient: Despite totaling just 0.8B parameters, this mixture outperforms every individual retriever and even larger 7B models by +10.8% and +3.9% on average, respectively. Further analysis also shows that this mixture framework can help incorporate specialized non-oracle human information sources as retrievers to achieve good collaboration, with a 58.9% relative performance improvement over simulated humans alone.

cs.IR↗

arXiv2025

Rigor in AI: Doing Rigorous AI Work Requires a Broader, Responsible AI-Informed Conception of Rigor

In AI research and practice, rigor remains largely understood in terms of methodological rigor -- such as whether mathematical, statistical, or computational methods are correctly applied. We argue that this narrow conception of rigor has contributed to the concerns raised by the responsible AI community, including overblown claims about the capabilities of AI systems. Our position is that a broader conception of what rigorous AI research and practice should entail is needed. We believe such a conception -- in addition to a more expansive understanding of (1) methodological rigor -- should include aspects related to (2) what background knowledge informs what to work on (epistemic rigor); (3) how disciplinary, community, or personal norms, standards, or beliefs influence the work (normative rigor); (4) how clearly articulated the theoretical constructs under use are (conceptual rigor); (5) what is reported and how (reporting rigor); and (6) how well-supported the inferences from existing evidence are (interpretative rigor). In doing so, we also provide useful language and a framework for much-needed dialogue about the AI community's work by researchers, policymakers, journalists, and other stakeholders.

cs.CY↗

arXiv2025

LTRR: Learning To Rank Retrievers for LLMs

Retrieval-Augmented Generation (RAG) systems typically rely on a single fixed retriever, despite growing evidence that no single retriever performs optimally across all query types. In this paper, we explore a query routing approach that dynamically selects from a pool of retrievers based on the query, using both train-free heuristics and learned routing models. We frame routing as a learning-to-rank problem and introduce LTRR, a framework that Learns To Rank Retrievers according to their expected contribution to downstream RAG performance. Through experiments on diverse question-answering benchmarks with controlled variations in query types, we demonstrate that routing-based RAG consistently surpasses the strongest single-retriever baselines. The gains are particularly substantial when training with the Answer Correctness (AC) objective and when using pairwise ranking methods, with XGBoost yielding the best results. Additionally, our approach exhibits stronger generalization to out-of-distribution queries. Overall, our results underscore the critical role of both training strategy and optimization metric choice in effective query routing for RAG systems.

cs.CL↗

arXiv2025

Contextual Metric Meta-Evaluation by Measuring Local Metric Accuracy

Meta-evaluation of automatic evaluation metrics -- assessing evaluation metrics themselves -- is crucial for accurately benchmarking natural language processing systems and has implications for scientific inquiry, production model development, and policy enforcement. While existing approaches to metric meta-evaluation focus on general statements about the absolute and relative quality of metrics across arbitrary system outputs, in practice, metrics are applied in highly contextual settings, often measuring the performance for a highly constrained set of system outputs. For example, we may only be interested in evaluating a specific model or class of models. We introduce a method for contextual metric meta-evaluation by comparing the local metric accuracy of evaluation metrics. Across translation, speech recognition, and ranking tasks, we demonstrate that the local metric accuracies vary both in absolute value and relative effectiveness as we shift across evaluation contexts. This observed variation highlights the importance of adopting context-specific metric evaluations over global ones.

cs.CL↗

arXiv2025

Tip of the Tongue Query Elicitation for Simulated Evaluation

Tip-of-the-tongue (TOT) search occurs when a user struggles to recall a specific identifier, such as a document title. While common, existing search systems often fail to effectively support TOT scenarios. Research on TOT retrieval is further constrained by the challenge of collecting queries, as current approaches rely heavily on community question-answering (CQA) websites, leading to labor-intensive evaluation and domain bias. To overcome these limitations, we introduce two methods for eliciting TOT queries - leveraging large language models (LLMs) and human participants - to facilitate simulated evaluations of TOT retrieval systems. Our LLM-based TOT user simulator generates synthetic TOT queries at scale, achieving high correlations with how CQA-based TOT queries rank TOT retrieval systems when tested in the Movie domain. Additionally, these synthetic queries exhibit high linguistic similarity to CQA-derived queries. For human-elicited queries, we developed an interface that uses visual stimuli to place participants in a TOT state, enabling the collection of natural queries. In the Movie domain, system rank correlation and linguistic similarity analyses confirm that human-elicited queries are both effective and closely resemble CQA-based queries. These approaches reduce reliance on CQA-based data collection while expanding coverage to underrepresented domains, such as Landmark and Person. LLM-elicited queries for the Movie, Landmark, and Person domains have been released as test queries in the TREC 2024 TOT track, with human-elicited queries scheduled for inclusion in the TREC 2025 TOT track. Additionally, we provide source code for synthetic query generation and the human query collection interface, along with curated visual stimuli used for eliciting TOT queries.

cs.IR↗

arXiv2024

Offline Evaluation of Set-Based Text-to-Image Generation

Text-to-Image (TTI) systems often support people during ideation, the early stages of a creative process when exposure to a broad set of relevant images can help explore the design space. Since ideation is an important subclass of TTI tasks, understanding how to quantitatively evaluate TTI systems according to how well they support ideation is crucial to promoting research and development for these users. However, existing evaluation metrics for TTI remain focused on distributional similarity metrics like Fr\'echet Inception Distance (FID). We take an alternative approach and, based on established methods from ranking evaluation, develop TTI evaluation metrics with explicit models of how users browse and interact with sets of spatially arranged generated images. Our proposed offline evaluation metrics for TTI not only capture how relevant generated images are with respect to the user's ideation need but also take into consideration the diversity and arrangement of the set of generated images. We analyze our proposed family of TTI metrics using human studies on image grids generated by three different TTI systems based on subsets of the widely used benchmarks such as MS-COCO captions and Localized Narratives as well as prompts used in naturalistic settings. Our results demonstrate that grounding metrics in how people use systems is an important and understudied area of benchmark design.

cs.CV↗