arXiv subjects

Thorsten Joachims

Publications and source records attributed to Thorsten Joachims.

At least 19 recordsLinked to original sources

arXiv2026

Value Over Language Model: Detecting Original Contribution in Writing

LLMs have been rapidly adopted across writing tasks, prompting the development of tools for detecting LLM-generated text. Yet, these tools largely measure how much of a document's surface text was written by an LLM and aren't fundamentally designed to measure how much of the information content or ideas originated from the LLM itself rather than being supplied by the user in the prompt. In this work, we design a framework that measures how much value a person adds on top of what a language model could have easily produced by itself. The method requires no training or labeled data and never scores the document's surface text, insulating it from stylistic confounders. Instead, it extracts the document's content at increasing levels of granularity, uses an LLM to reconstruct the document from each partial representation, and compares these reconstructions with those produced from the task description alone. We call this framework Value Over Language Model (VOLM), which measures a document's contribution relative to a replacement-level document that an LLM could produce from the task description alone. We evaluate VOLM with a specific instantiation of this framework across three domains: news articles, ICLR peer reviews, and argumentative essays. VOLM separates human-authored documents from matched LLM-generated documents produced from generic task descriptions, while remaining substantially invariant to content-preserving transformations, including LLM-based reconstruction and round-trip translation. We further find that increasingly constrained content extractors reduce residual differences between LLM-generated and humanized text, demonstrating the importance of disentangling informational content from stylistic variation. We hope these results encourage further work on specialized instantiations of the framework and on assessing human contributions in LLM-assisted writing more generally.

cs.AI↗

arXiv2026

Credit-assigned Policy Gradient for Early Stage Retrieval in Two-stage Ranking

Large-scale search, recommendation, and retrieval-augmented generation (RAG) systems typically employ a two-stage architecture: an early-stage ranker (ESR) generates a candidate set, which is subsequently re-ranked by a late-stage ranker (LSR). While there are many reinforcement learning (RL) methods for training the LSR, end-to-end training of the ESR has proven challenging. In particular, naive application of "vanilla" policy gradient (V-PG) is not scalable for candidate-set sizes relevant for practical use due to exploding variance. This issue arises because V-PG propagates the gradient to the joint probability of the candidate sets, ignoring the contribution of each specific item in the candidate set to the reward. To mitigate this issue, we propose a novel "credit-assigned" policy gradient (CA-PG), which computes gradients with respect to the probability that the target item is chosen in any candidate set, i.e. marginalizing over all candidate sets that contain it. Our theoretical analysis reveals that CA-PG significantly reduces the variance of V-PG by marginalizing over the specific composition of the candidate set, while preserving the ability to learn the correct ranking of items under a reasonably aligned LSR policy. Experiments on both synthetic and real-world data demonstrate that CA-PG improves the convergence speed and training stability for ESRs utilizing the canonical Plackett-Luce model, especially when the candidate-set size is large.

cs.IR↗

arXiv2026

$p1$: Better Prompt Optimization with Fewer Prompts

Prompt optimization improves language models without updating their weights by searching for a better system prompt, but its effectiveness varies widely across tasks. We study what makes a task amenable to prompt optimization. We show that the reward variance across different system prompts can be decomposed into two components: variance among responses, which captures generation stochasticity, and variance among system prompts, which captures differences in system prompt quality. Prompt optimization succeeds when variance among system prompts is sufficiently large, but fails when variance among responses dominates the variance of the system prompts. Surprisingly, we further show that scaling to more user prompts can hurt optimization by reducing variance among system prompts, especially on heterogeneous datasets where different user prompts favor different system prompts. Motivated by this insight, we propose $p1$, a simple user prompt filtering method that selects a small subset of user prompts with high variance across candidate system prompts. This subset of user prompts allows one to distinguish a good system prompt from a bad one, making system optimization easier. Experiments on reasoning benchmarks show that $p1$ substantially improves prompt optimization over training on the full dataset and outperforms strong baselines such as GEPA. Notably, training on only two prompts from AIME 24 yields a system prompt that generalizes well to other reasoning benchmarks.

cs.LG↗

arXiv2026

The Digital Divide in Generative AI: Evidence from Large Language Model Use in College Admissions Essays

Large language models (LLMs) have become popular writing tools among students and may expand access to high-quality feedback for students with less access to traditional writing support. At the same time, LLMs may standardize student voice or invite overreliance. This study examines how adoption of LLM-assisted writing varies across socioeconomic groups and how it relates to outcomes in a high-stakes context: U.S. college admissions. We analyze a de-identified longitudinal dataset of applications to a selective university from 2020 to 2024 (N = 81,663). Estimating LLM use using a distribution-based detector trained on synthetic and historical essays, we tracked how student writing changed as LLM use proliferated, how adoption differed by socioeconomic status (SES), and whether potential benefits translated equitably into admissions outcomes. Using fee-waiver status as a proxy for SES, we observe post-2023 convergence in surface-level linguistic features, with the largest changes in fee-waived and rejected applicants. Estimated LLM use rose sharply in 2024 across all groups, with disproportionately larger increases among lower SES applicants, consistent with an access hypothesis in which LLMs substitute for scarce writing support. However, increased estimated LLM use was more strongly associated with declines in predicted admission probability for lower SES applicants than for higher SES applicants, even after controlling for academic credentials and stylometric features. These findings raise concerns about equity and the validity of essay-based evaluation in an era of AI-assisted writing and provide the first large-scale longitudinal evidence linking LLM adoption, linguistic change, and evaluative outcomes in college admissions.

cs.CY↗

arXiv2026

Does Algorithmic Uncertainty Sway Human Experts? Evidence from a Field Experiment in Selective College Admissions

Algorithmic predictions are inherently uncertain: even models with similar aggregate accuracy can produce different predictions for the same individual, raising concerns that high-stakes decisions may become sensitive to arbitrary modeling choices. In this paper, we define \emph{algorithmic sensitivity} as the extent to which arbitrary modeling choices propagate into human decisions: how much a decision outcome shifts when a more favorable versus less favorable algorithmic prediction is presented to the decision-maker for the same individual. We estimate this in a randomized field experiment ($n=19{,}545$) embedded in a selective U.S. college admissions cycle, in which admissions officers reviewed each application alongside an algorithmic score while we randomly varied whether the score came from one of two similarly accurate prediction models. Although the two models performed similarly in aggregate, they frequently assigned different scores to the same applicant, creating exogenous variation in the score shown. Surprisingly, we find little evidence of algorithmic sensitivity: presenting a more favorable score does not meaningfully increase an applicant's probability of admission on average, even when the models disagree substantially. These findings suggest that, in this expert, high-stakes setting, human decision-making is largely invariant to arbitrary variation in algorithmic predictions, underscoring the role of professional discretion and institutional context in mediating the downstream effects of algorithmic uncertainty.

cs.HC↗

arXiv2026

Do LLMs Favor LLMs? Quantifying Interaction Effects in Peer Review

There are increasing indications that LLMs are not only used for producing scientific papers, but also as part of the peer review process. In this work, we provide the first comprehensive analysis of LLM use across the peer review pipeline, with particular attention to interaction effects: not just whether LLM-assisted papers or LLM-assisted reviews are different in isolation, but whether LLM-assisted reviews evaluate LLM-assisted papers differently. In particular, we analyze over 125,000 paper-review pairs from ICLR, NeurIPS, and ICML. We initially observe what appears to be a systematic interaction effect: LLM-assisted reviews seem especially kind to LLM-assisted papers compared to papers with minimal LLM use. However, controlling for paper quality reveals a different story: LLM-assisted reviews are simply more lenient toward lower quality papers in general, and the over-representation of LLM-assisted papers among weaker submissions creates a spurious interaction effect rather than genuine preferential treatment of LLM-generated content. By augmenting our observational findings with reviews that are fully LLM-generated, we find that fully LLM-generated reviews exhibit severe rating compression that fails to discriminate paper quality, while human reviewers using LLMs substantially reduce this leniency. Finally, examining metareviews, we find that LLM-assisted metareviews are more likely to render accept decisions than human metareviews given equivalent reviewer scores, though fully LLM-generated metareviews tend to be harsher. This suggests that meta-reviewers do not merely outsource the decision-making to the LLM. These findings provide important input for developing policies that govern the use of LLMs during peer review, and they more generally indicate how LLMs interact with existing decision-making processes.

cs.AI↗

arXiv2026

SteerEval: A Framework for Evaluating Steerability with Natural Language Profiles for Recommendation

Natural-language user profiles have recently attracted attention not only for improved interpretability, but also for their potential to make recommender systems more steerable. By enabling direct editing, natural-language profiles allow users to explicitly articulate preferences that may be difficult to infer from past behavior. However, it remains unclear whether current natural-language-based recommendation methods can follow such steering commands. While existing steerability evaluations have shown some success for well-recognized item attributes (e.g., movie genres), we argue that these benchmarks fail to capture the richer forms of user control that motivate steerable recommendations. To address this gap, we introduce SteerEval, an evaluation framework designed to measure more nuanced and diverse forms of steerability by using interventions that range from genres to content-warning for movies. We assess the steerability of a family of pretrained natural-language recommenders, examine the potential and limitations of steering on relatively niche topics, and compare how different profile and recommendation interventions impact steering effectiveness. Finally, we offer practical design suggestions informed by our findings and discuss future steps in steerable recommender design.

cs.IR↗

arXiv2025

Prompt Curriculum Learning for Efficient LLM Post-Training

We introduce Prompt Curriculum Learning (PCL), a lightweight reinforcement learning (RL) algorithm that selects intermediate-difficulty prompts using a learned value model to post-train language models. Since post-training LLMs via RL remains sensitive to batching and prompt selection strategies, we first conduct a series of systematic experiments where we (1) determine the optimal training batch size that balances generation efficiency and gradient quality and (2) establish the importance of focusing on prompts of intermediate difficulty for the policy. We build upon these results to design PCL, which identifies prompts of intermediate difficulty for the current policy in an on-policy manner by using a value model that is concurrently updated based on the current policy. By focusing on informative prompts that yield high effective ratios, PCL achieves either the highest performance or requires significantly less time to reach comparable performance to its counterparts. Compared to rollout-based filtering methods, PCL avoids costly rollouts and achieves $12.1\times$ and $16.9\times$ faster speed on identifying intermediate-difficulty prompts when training on MATH and DeepScaleR, respectively. We further demonstrate that our value model accurately predicts prompt difficulty and allows PCL to focus on progressively more challenging prompts during RL. Our results present a new methodology that delivers improved tradeoff between upper-bound performance and efficiency for reasoning-focused RL.

cs.LG↗

arXiv2025

Prompt Optimization with Logged Bandit Data

We study how to use naturally available user feedback, such as clicks, to optimize large language model (LLM) pipelines for generating personalized sentences using prompts. Naive approaches, which estimate the policy gradient in the prompt space, suffer either from variance caused by the large action space of prompts or bias caused by inaccurate reward predictions. To circumvent these challenges, we propose a novel kernel-based off-policy gradient method, which estimates the policy gradient by leveraging similarity among generated sentences, substantially reducing variance while suppressing the bias. Empirical results on our newly established suite of benchmarks demonstrate the effectiveness of the proposed approach in generating personalized descriptions for movie recommendations, particularly when the number of candidate prompts is large.

cs.LG↗

arXiv2025

Poor Alignment and Steerability of Large Language Models: Evidence from College Admission Essays

People are increasingly using technologies equipped with large language models (LLM) to write texts for formal communication, which raises two important questions at the intersection of technology and society: Who do LLMs write like (model alignment); and can LLMs be prompted to change who they write like (model steerability). We investigate these questions in the high-stakes context of undergraduate admissions at a selective university by comparing lexical and sentence variation between essays written by 30,000 applicants to two types of LLM-generated essays: one prompted with only the essay question used by the human applicants; and another with additional demographic information about each applicant. We consistently find that both types of LLM-generated essays are linguistically distinct from human-authored essays, regardless of the specific model and analytical approach. Further, prompting a specific sociodemographic identity is remarkably ineffective in aligning the model with the linguistic patterns observed in human writing from this identity group. This holds along the key dimensions of sex, race, first-generation status, and geographic location. The demographically prompted and unprompted synthetic texts were also more similar to each other than to the human text, meaning that prompting did not alleviate homogenization. These issues of model alignment and steerability in current LLMs raise concerns about the use of LLMs in high-stakes contexts.

cs.CL↗

arXiv2025

MultiScale Contextual Bandits for Long Term Objectives

The feedback that AI systems (e.g., recommender systems, chatbots) collect from user interactions is a crucial source of training data. While short-term feedback (e.g., clicks, engagement) is widely used for training, there is ample evidence that optimizing short-term feedback does not necessarily achieve the desired long-term objectives. Unfortunately, directly optimizing for long-term objectives is challenging, and we identify the disconnect in the timescales of short-term interventions (e.g., rankings) and the long-term feedback (e.g., user retention) as one of the key obstacles. To overcome this disconnect, we introduce the framework of MultiScale Policy Learning to contextually reconcile that AI systems need to act and optimize feedback at multiple interdependent timescales. Following a PAC-Bayes motivation, we show how the lower timescales with more plentiful data can provide a data-dependent hierarchical prior for faster learning at higher scales, where data is more scarce. As a result, the policies at all levels effectively optimize for the long-term. We instantiate the framework with MultiScale Off-Policy Bandit Learning (MSBL) and demonstrate its effectiveness on three tasks relating to recommender and conversational systems.

cs.LG↗

arXiv2024

End-to-end Training for Recommendation with Language-based User Profiles

There is a growing interest in natural language-based user profiles for recommender systems, which aims to enhance transparency and scrutability compared with embedding-based methods. Existing studies primarily generate these profiles using zero-shot inference from large language models (LLMs), but their quality remains insufficient, leading to suboptimal recommendation performance. In this paper, we introduce LangPTune, the first end-to-end training framework to optimize LLM-generated user profiles. Our method significantly outperforms zero-shot approaches by explicitly training the LLM for the recommendation objective. Through extensive evaluations across diverse training configurations and benchmarks, we demonstrate that LangPTune not only surpasses zero-shot baselines but can also matches the performance of state-of-the-art embedding-based methods. Finally, we investigate whether the training procedure preserves the interpretability of these profiles compared to zero-shot inference through both GPT-4 simulations and crowdworker user studies. Implementation of LangPTune can be found at https://github.com/ZhaolinGao/LangPTune.

cs.IR↗

arXiv2024

Algorithms for College Admissions Decision Support: Impacts of Policy Change and Inherent Variability

Each year, selective American colleges sort through tens of thousands of applications to identify a first-year class that displays both academic merit and diversity. In the 2023-2024 admissions cycle, these colleges faced unprecedented challenges. First, the number of applications has been steadily growing. Second, test-optional policies that have remained in place since the COVID-19 pandemic limit access to key information historically predictive of academic success. Most recently, longstanding debates over affirmative action culminated in the Supreme Court banning race-conscious admissions. Colleges have explored machine learning (ML) models to address the issues of scale and missing test scores, often via ranking algorithms intended to focus on 'top' applicants. However, the Court's ruling will force changes to these models, which were able to consider race as a factor in ranking. There is currently a poor understanding of how these mandated changes will shape applicant ranking algorithms, and, by extension, admitted classes. We seek to address this by quantifying the impact of different admission policies on the applications prioritized for review. We show that removing race data from a developed applicant ranking algorithm reduces the diversity of the top-ranked pool without meaningfully increasing the academic merit of that pool. We contextualize this impact by showing that excluding data on applicant race has a greater impact than excluding other potentially informative variables like intended majors. Finally, we measure the impact of policy change on individuals by comparing the arbitrariness in applicant rank attributable to policy change to the arbitrariness attributable to randomness. We find that any given policy has a high degree of arbitrariness and that removing race data from the ranking algorithm increases arbitrariness in outcomes for most applicants.

cs.CY↗

arXiv2024

REBEL: Reinforcement Learning via Regressing Relative Rewards

While originally developed for continuous control problems, Proximal Policy Optimization (PPO) has emerged as the work-horse of a variety of reinforcement learning (RL) applications, including the fine-tuning of generative models. Unfortunately, PPO requires multiple heuristics to enable stable convergence (e.g. value networks, clipping), and is notorious for its sensitivity to the precise implementation of these components. In response, we take a step back and ask what a minimalist RL algorithm for the era of generative models would look like. We propose REBEL, an algorithm that cleanly reduces the problem of policy optimization to regressing the relative reward between two completions to a prompt in terms of the policy, enabling strikingly lightweight implementation. In theory, we prove that fundamental RL algorithms like Natural Policy Gradient can be seen as variants of REBEL, which allows us to match the strongest known theoretical guarantees in terms of convergence and sample complexity in the RL literature. REBEL can also cleanly incorporate offline data and be extended to handle the intransitive preferences we frequently see in practice. Empirically, we find that REBEL provides a unified approach to language modeling and image generation with stronger or similar performance as PPO and DPO, all while being simpler to implement and more computationally efficient than PPO. When fine-tuning Llama-3-8B-Instruct, REBEL achieves strong performance in AlpacaEval 2.0, MT-Bench, and Open LLM Leaderboard.

cs.LG↗

arXiv2024

Language-Based User Profiles for Recommendation

Most conventional recommendation methods (e.g., matrix factorization) represent user profiles as high-dimensional vectors. Unfortunately, these vectors lack interpretability and steerability, and often perform poorly in cold-start settings. To address these shortcomings, we explore the use of user profiles that are represented as human-readable text. We propose the Language-based Factorization Model (LFM), which is essentially an encoder/decoder model where both the encoder and the decoder are large language models (LLMs). The encoder LLM generates a compact natural-language profile of the user's interests from the user's rating history. The decoder LLM uses this summary profile to complete predictive downstream tasks. We evaluate our LFM approach on the MovieLens dataset, comparing it against matrix factorization and an LLM model that directly predicts from the user's rating history. In cold-start settings, we find that our method can have higher accuracy than matrix factorization. Furthermore, we find that generating a compact and human-readable summary often performs comparably with or better than direct LLM prediction, while enjoying better interpretability and shorter model input length. Our results motivate a number of future research directions and potential improvements.

cs.CL↗

arXiv2024

Reviewer2: Optimizing Review Generation Through Prompt Generation

Recent developments in LLMs offer new opportunities for assisting authors in improving their work. In this paper, we envision a use case where authors can receive LLM-generated reviews that uncover weak points in the current draft. While initial methods for automated review generation already exist, these methods tend to produce reviews that lack detail, and they do not cover the range of opinions that human reviewers produce. To address this shortcoming, we propose an efficient two-stage review generation framework called Reviewer2. Unlike prior work, this approach explicitly models the distribution of possible aspects that the review may address. We show that this leads to more detailed reviews that better cover the range of aspects that human reviewers identify in the draft. As part of the research, we generate a large-scale review dataset of 27k papers and 99k reviews that we annotate with aspect prompts, which we make available as a resource for future research.

cs.CL↗

arXiv2024

POTEC: Off-Policy Learning for Large Action Spaces via Two-Stage Policy Decomposition

We study off-policy learning (OPL) of contextual bandit policies in large discrete action spaces where existing methods -- most of which rely crucially on reward-regression models or importance-weighted policy gradients -- fail due to excessive bias or variance. To overcome these issues in OPL, we propose a novel two-stage algorithm, called Policy Optimization via Two-Stage Policy Decomposition (POTEC). It leverages clustering in the action space and learns two different policies via policy- and regression-based approaches, respectively. In particular, we derive a novel low-variance gradient estimator that enables to learn a first-stage policy for cluster selection efficiently via a policy-based approach. To select a specific action within the cluster sampled by the first-stage policy, POTEC uses a second-stage policy derived from a regression-based approach within each cluster. We show that a local correctness condition, which only requires that the regression model preserves the relative expected reward differences of the actions within each cluster, ensures that our policy-gradient estimator is unbiased and the second-stage policy is optimal. We also show that POTEC provides a strict generalization of policy- and regression-based approaches and their associated assumptions. Comprehensive experiments demonstrate that POTEC provides substantial improvements in OPL effectiveness particularly in large and structured action spaces.

stat.ML↗

arXiv2024

Ranking with Long-Term Constraints

The feedback that users provide through their choices (e.g., clicks, purchases) is one of the most common types of data readily available for training search and recommendation algorithms. However, myopically training systems based on choice data may only improve short-term engagement, but not the long-term sustainability of the platform and the long-term benefits to its users, content providers, and other stakeholders. In this paper, we thus develop a new framework in which decision makers (e.g., platform operators, regulators, users) can express long-term goals for the behavior of the platform (e.g., fairness, revenue distribution, legal requirements). These goals take the form of exposure or impact targets that go well beyond individual sessions, and we provide new control-based algorithms to achieve these goals. In particular, the controllers are designed to achieve the stated long-term goals with minimum impact on short-term engagement. Beyond the principled theoretical derivation of the controllers, we evaluate the algorithms on both synthetic and real-world data. While all controllers perform well, we find that they provide interesting trade-offs in efficiency, robustness, and the ability to plan ahead.

cs.IR↗