paperlist.ai

AI conference papers

AI conference paper library

Search papers from 30+ top conferences, including NeurIPS, ICLR, ICML, CVPR, ACL and AAAI. Filter by keyword, then open any paper for its abstract, authors and links.

arXiv

Beyond Static and Linear: What Attention Constraints Best Fit Human Reading Times?

Lanni Bu, Xiulin Yang, Christian Clark, Alex Warstadt, Ethan Gotlieb Wilcox

Transformer-based language models are widely used as models of human language processing, yet their attention mechanisms allow lossless access to the full preceding context, unlike the limited memory systems of humans. We hypothesize that installing memory constraints into transformers' attention mechanisms can improve their fit to human behavioral data. While previous work has explored individual constraints in isolation, we conduct a systematic comparison of multiple attention-based memory mechanisms across different model sizes and training corpora, evaluating both psychometric predictive power for human reading times and grammatical competence. We additionally compare static constraints, in which the constraint strength is fixed throughout training, to dynamic memory curricula. We find that constraints that are sensitive to the content of intervening tokens consistently achieve the highest alignment with human reading times, outperforming distance-based constraints. We observe a dissociation between psychometric fit and grammatical competence under dynamic memory curricula, suggesting that Transformers cannot serve as a one-size-fits-all cognitive model.

arXiv

Learning to Grade Efficiently: A Bandit-Driven Prompt-Selection Framework for Low-Cost LLM Essay Scoring

Olga Manakina, Igor Bogdanov

Large Language Models (LLMs) demonstrate strong capabilities in automated essay scoring (AES), but contemporary approaches typically employ fixed prompt selection, failing to address operational cost concerns and evolving optimal configurations. We propose a novel cost-aware approach that treats each prompt type as an arm in a multi-armed bandit (MAB) controller, enabling adaptive selection of optimal prompting strategies during inference. Our experiments on IELTS Writing Task 2 essays show that the MAB framework achieves comparable scoring accuracy to exhaustive grid search while reducing LLM calls by 78.4\% to find the best grading approach. We implemented four distinct grading recipes (multi-step vs. single-step assessment, with vs. without calibration examples) and found that the multi-step approach with examples achieves the highest accuracy. By tracking token usage and latency alongside agreement metrics, we produce the first cost-reliability learning curves for essay scoring, providing actionable insights for educational technology platforms that must balance operational costs against assessment validity. This work represents the first application of online control mechanisms to adaptively select prompting strategies in AES, transforming prompt selection from an offline hyperparameter optimization problem into an efficient online learning task.

arXiv

From Preferences to Principles: Rubric-Based Alignment for Grounded Knowledge Answers

Aman Saini, Priyanshu Kumar, Eric Peng, Kai Yuan, Harsh Girase, Wanming Chen

Designing effective reward signals for open-domain question answering is challenging because high-quality responses must simultaneously satisfy multiple aspects of answer quality that are difficult to capture with a holistic scalar objective. We introduce a rubric-based reward framework that generates query-specific rubrics grounded in retrieved evidence and decomposed into multiple quality dimensions, providing fine-grained supervision during post-training. Averaged across three evaluation axes (composition, grounding, and instruction-following), our approach improves over the instruction-tuned baseline by 6.5% and over flat rubric variants by 4%, with consistent gains across all evaluation datasets. Conditioning rubrics on retrieved evidence improves factual support, while decomposing rubrics into quality-specific dimensions further improves coherence, organization, and adherence to query requirements. Our results show that grounded, multi-dimensional rubrics provide more effective reward supervision for complex open-domain question answering.

arXiv

Discovering Cross-Language Reasoning Invariance in LLMs with Geometry-Invariant Sparse Autoencoders

Igor Bogdanov, Changcheng Huang

Multilingual language models can solve the same mathematical problem in different languages, but it remains unclear whether they rely on shared features or on language-specific computations that only produce similar outputs. We study this question in five models from four families using the Multilingual Grade School Math (MGSM) dataset, with problems solved in English, German, French, Spanish, Russian, and Chinese, retaining problems with valid reasoning traces in all six languages and replaying those traces through the model to record representations at multiple layers. For each model, we first use Centered Kernel Alignment (CKA) to identify layers with cross-language alignment. At each selected layer, we train two sparse autoencoders (SAE): a baseline reconstruction-only model and a contrastive variant introduced in this work, the Geometry-Invariant SAE (GI-SAE). GI-SAE supplements the reconstruction loss with an Information Noise-Contrastive Estimation (InfoNCE) loss that trains the encoder to produce similar activations for traces of the same problem, regardless of language or token position. We then test whether the resulting shared features are functionally interchangeable by swapping their values between languages during the model's forward pass and measuring the resulting change in output, quantified by Kullback-Leibler (KL) divergence per feature. Although GI-SAE yields higher CKA and Jaccard similarity at nearly every layer, higher geometric similarity does not consistently imply greater functional interchangeability. We find that cross-language feature sharing is model- and architecture-dependent in this sample and appears at different depths in different models. GI-SAE primarily amplifies cross-language structure already present: the pattern is model-specific, with strengthening in Qwen, no functional benefit in Gemma, and mixed layer-dependent effects in Llama and Phi.

arXiv

Giga-Embeddings: Mixture-of-Experts Encoders for High-Throughput Text Embeddings

Egor Kolodin, Egor Krasnoperov, Evgeniy Kosarev, Fyodor Minkin

We introduce Giga-Embeddings, a family of text embedding models designed to combine strong retrieval quality with efficient serving. Its largest member is a sparse 10B-parameter Mixture-of-Experts encoder with approximately 1.8B active parameters per token. Across English, Russian, multilingual, and code MTEB benchmarks, this model achieves the strongest aggregate performance within the family on all four evaluated suites. In our vLLM benchmark with 1024-token inputs, it processes 114.5k tokens per second, providing 25 percent higher throughput than the dense 3B model and 1.56-2.65x the throughput of the evaluated external systems. The family also includes a dense 3B encoder and a distilled 480M encoder for tighter compute and memory budgets. We train the compact model using a dimension-agnostic objective that aligns teacher and student similarity distributions. The resulting 480M model scores 70.98 on Russian MTEB, surpassing FRIDA while using 42 percent fewer parameters. We release all three model checkpoints.

arXiv

Inter-dimension Dependence for Multi-Dimensional Evaluation of Open-Ended Text

Haoyuan Li, Snigdha Chaturvedi

LLM-as-a-judge methods are widely used for evaluating the quality of generated open-ended text. Such evaluations are generally multi-dimensional, since the error patterns in texts can be different for different dimensions. Therefore, reliable LLM judges should evaluate each target dimension independently. To quantify the extent to which LLM judges depend on non-target dimensions when evaluating a target dimension, i.e., inter-dimension dependence, we propose CorrGap. To measure this, CorrGap uses the difference in correlations between LLM-predicted scores and ground truth scores across different groups of texts. Using CorrGap, we show that inter-dimension dependence is pervasive across LLM judges in open-ended text evaluation tasks. To mitigate inter-dimension dependence, we propose DimCheck, a method that iteratively removes unrelated evidence from COTs generated by LLM judges in a step-wise way. We show that DimCheck mitigates inter-dimension dependence and outperforms strong baselines across three LLMs and four tasks. We also show that smaller trained LLMs can approximate larger LLMs in DimCheck, with much lower inference costs.

arXiv

What Reaches Expert Review? Representation, Structural Screening, and Candidate-Form Dependence in AI-Assisted Item Development

Christopher Brooks

Between AI-assisted item generation and expert review sits a computational evaluator whose decisions are usually treated as technical preliminaries. Yet representation, structural reduction, and selection policy determine which items and evidence psychometricians ever receive. Across two linked in-silico studies of 32,000 selected Big Five items, we followed fixed source populations from semantic representation through structural evaluation and candidate-form construction. Broad agreement in semantic geometry concealed consequential local differences: identical wording acquired different construct evidence, different items survived, and intended attributes could disappear even as community correspondence improved. These sensitivities also differed across generated source populations. At the final review boundary, both eligibility policies filled every content cell in every evaluable form, yet they presented different wording. Across embedding configurations, inclusive primary forms shared a median of only 6 of 40 items, reflecting the total downstream consequence of changing representation across structural evidence and ranking. The apparent stability of global summaries and complete forms therefore concealed instability in the content reaching psychometricians. The computational evaluator is not neutral infrastructure between generation and expertise; it is an inspectable and revisable part of measurement design.

arXiv

Calibration-Preserving Pruning: Compression as a Reliability Contract

Ibne Farabi Shihab, Adria Binte Habib, Anuj Sharma

Split conformal prediction, not the pruning rule, supplies finite-sample marginal coverage once a pruned model is fixed independently of the conformal calibration split. We study the separate efficiency problem: can pruning preserve score geometry well enough to obtain smaller valid prediction sets? Calibration-Preserving Pruning (CPP) augments a base pruning score with nonconformity-gradient saliency and uses disjoint pruning, validation-selection, conformal-calibration, and test splits. Bounded score perturbations imply bounded conformal-quantile shifts and controlled set inflation, but do not make the generic coverage theorem CPP-specific. Final five-seed Qwen2.5-1.5B results at 50\% sparsity show the largest gains on large-label tasks. On DBpedia-14, CPP-SparseGPT reduces mean set size from \(10.1\) to \(8.6\) while changing accuracy from \(0.347\) to \(0.366\); CPP-Wanda reduces \(11.2\) to \(9.0\) with an accuracy trade-off from \(0.310\) to \(0.295\). Across 15 dataset--sparsity cells, CPP-SparseGPT produces smaller sets in 13 and higher accuracy in 11. Matched controls show that generic supervised gradients explain much of the gain: true-label CPP is not statistically resolved from matched Wanda+SNIP, whereas threshold-aware candidate-label CPP reaches \(7.8\) mean set size at explicit accuracy and offline-compute costs. RoBERTa-base and Llama-3-8B diagnostics support transfer, but our claims remain limited to reliability-sensitive classification.

arXiv

ADE: Agentic Data Evolution Framework for Human-Centered Objectives

Yang Yu, Yilin Jiang, Zexuan Fei, Yiming Luo, Xingkai Song, Kaiyi Huang, Aimin Zhou, Xin Lin, Fei Tan

Aligning large language models to human-centered objectives is difficult when targets are non-executable and context-dependent, limiting reliable verification and scalable supervision. Although synthetic data expands coverage, weak verification shifts the bottleneck from generation to selection. Noisy signals destabilize iterative refinement and can cause silent regressions. We propose Agentic Data Evolution (ADE), a data-centric framework that organizes synthetic supervision as evolving data snapshots. ADE improves data snapshots through a closed-loop Observation-Variation-Selection (OVS) procedure, where a steady-state admission mechanism acts as a quality ratchet that conservatively gates updates for sustained cross-round improvement. We validate these improvements through complementary intrinsic trend tracking and extrinsic post-training evaluation. On DEV300, ADE raises the intrinsic win rate from 50% to 75.81% and the extrinsic win rate from 55.20% to 68.86%, consistent performance gains across diverse benchmarks. Blind expert evaluation further confirms this, with a 66.11% preference for evolved answers. These gains extend across post-training methods, model scales, and tasks beyond the target weakly verifiable educational objectives. Resources are available at https://github.com/ZeroLoss-Lab/Agentic-Data-Evolution.

arXiv

Automata from Agent Traces: Failure and Next-Step Prediction

Seonglae Cho, Franklin Cardenoso Fernandez, Umar Mohammed, Zekun Wu, Kleyton Da Costa, Ilham Wicaksono, Adriano Koshiyama

LLM-based agents execute multi-step tasks, but their behavioral structure remains opaque: long unstructured traces resist the safety auditing and runtime monitoring that deployment requires. Existing approaches operate per-trace or success-only, so they miss the cross-run topology that links next-step and failure prediction. To recover that shared structure, we collapse an entire trace corpus into a single, compact finite-state machine (FSM) that serves as a structural substrate for the otherwise unpredictable behavior of LLM agents. Across twelve public datasets, the FSMs are compact (7-43 states), replay held-out data at >=0.997 fitness with near-identical topology across splits, and build in milliseconds. This substrate addresses both prediction goals. For next-step prediction, FSM-state context outperforms Agent Workflow Memory on every ground-truth-matched dataset. For failure prediction, per-state behavioral features reach held-out AUROC up to 0.94, and an online monitor ranks failing runs above passing ones from a partial trace, triggering early stopping well before completion. Behavioral topology thus appears shaped more by the deployment harness than by the LLM, providing a model-agnostic structural primitive for safety auditing and runtime monitoring.

arXiv

Gated Activation Steering for Reducing Sycophancy & Hallucination in Medical Question Answering

Himanshu Tripathi, Subash Neupane, Shaswata Mitra, Sudip Mittal, Noorbakhsh Amiri Golilarz, Shahram Rahimi

Sycophancy and hallucination are persistent failure modes of Large Language Models (LLMs) across domains. However, it becomes particularly consequential in clinical question answering, where responses must remain grounded in the provided context and robust to user pressure. Hallucination can introduce information that is unsupported by the context, while sycophancy can cause a model to abandon a previously correct answer when challenged by the user. Existing approaches, such as prompt-based safeguards and always-on activation steering, often address these behaviors separately or apply interventions broadly across turns, which can unnecessarily deteriorate responses that were already correct. To address these limitations within a single framework, we employ Inference Time Intervention (ITI) to jointly control both behaviors by learning separate steering directions for hallucination and sycophancy from contrastive clinical pairs and applying them to causally verified attention heads. During runtime, behavior-specific gates then determine when intervention is needed: the hallucination component mitigates unsupported claims, while the sycophancy component mitigates answer shifts caused by user pressure. We evaluate this framework on clinical questions grounded in EHR data while keeping the model weights frozen. Across all evaluation settings, we conducted 15,900 model-response runs. Across 600 pressure trajectories for the 4-billion-parameter model, the unsteered model caved in 570 cases. At the same time, gated steering helped it last longer in 551 of them. It held its ground under pressure at levels comparable to those of models with more than 100 billion parameters, showing that targeted inference-time steering can improve robustness without intervening at every turn.

arXiv

Contextual Embedding Evidence for Main--Light Verb Distinctions in Urdu

Farah Adeeba, Miriam Butt

Urdu light verbs contribute schematic event-structural meaning while remaining lexically related to corresponding main verbs. This study tests representational predictions derived from Butt's analysis using contextual embeddings from UrduBERT, DunbaaBERT, and multilingual BERT across 1,126 naturally occurring sentences containing seven Urdu verbs. Main and light uses show significant representational separation in all 21 verb--model comparisons. At the same time, same-lemma main and light centroids are consistently closer than mismatched main--light lemma pairs, supporting continued lexical relatedness. In a seven-way prediction task restricted to light uses, verb identity remains recoverable after the target is masked, with UrduBERT achieving 0.866 accuracy and 0.852 macro-F1. UrduBERT also retains 0.782 accuracy under a preceding-form-disjoint evaluation, indicating generalization beyond repeated local verb combinations. These findings provide computational evidence consistent with Butt's account that Urdu light verbs differ systematically from their main uses while retaining lemma-specific and verb-specific representational structure.

arXiv

Tool-Making and Self-Evolving LLM Agents in Low-Latency Systems

Kalle Kujanpää, Ning Liu, Shahnawaz Alam, Yeshwanth Reddy Sura, Tianyu Yang, Kristina Klinkner, Shervin Malmasi

Production LLM agents often waste latency and reliability by regenerating code for the same procedural steps on every request. We replace this inference-time coding loop with an agentic tool-making pipeline that compiles repeated SOP steps into validated, versioned tools before deployment. The tool-maker grounds synthesis in the live environment as it collects execution traces, observes backend schemas and values, generates candidate tools, and repairs them against labeled cases. At runtime, the production agent calls these tools directly and falls back to code generation only when needed. We deploy the approach in a Fulfillment Center alarm-triage system, where an agent diagnoses alarms against a 44-node SOP over heterogeneous metric backends. In production, tool calls reduce p50 latency by 42%. On 1,500 historical alarms, they reduce end-to-end error rate by up to 53% by suppressing run-to-run variance in repeated steps. Because tools return compact structured verdicts, they also enable a simpler direct-call architecture, reducing p50 latency by a further 62% in a controlled ablation. Versioned tools also improve auditability and expose specification gaps and upstream data drift. Our results show that agents that build and maintain their own tool libraries, a key element of self-evolving agents, can make industrial LLM systems faster, more reliable, and easier to operate.

arXiv

Don't Commit Alone: Joint Token Commitment in Diffusion Language Models

Lin Yao

Diffusion language models (dLLMs) commit multiple tokens per denoising step by decoding each selected position independently from a shared context. When these positions are dependent, this factorization introduces an error captured by conditional total correlation, which confidence-based selection cannot infer from marginal probabilities alone. We propose CoCommit, a marker-gated coordination pass that delays commitment. After the usual bundle selection, a learned marker identifies the commit set, and the backbone's last n layers are re-applied to coordinate the marked positions before greedy argmax writes the tokens. This approximates joint-mode decoding while reusing existing weights, requiring only one partial forward pass and no auxiliary model. On LLaDA 2.1 with LoRA adapters and greedy inference, joint commitment improves five of the seven evaluated benchmarks over the released factorized decoder. The largest gains occur on code and reasoning tasks, while the remaining tasks are near parity.

arXiv

SaliMory: Orchestrating Cognitive Memory for Conversational Agents

Kai Zhang, Xinyuan Zhang, Hongda Jiang, Shiun-Zu Kuo, Hyokun Yun, Ejaz Ahmed, Shereen Oraby, Ziyun Li, Sanat Sharma, Ann Lee, Ahmed A Aly, Anuj Kumar, Raffay Hamid, Xin Luna Dong

Conversational agents that serve as lifelong companions must maintain persistent memory across all interactions. However, simply expanding context windows with raw retrieval degrades reasoning quality, while training memory agents via standard reinforcement learning creates a severe credit assignment bottleneck in a multi-stage pipeline. To solve this, we introduce SALIMORY, a framework that trains a single language model to manage a cognitively-structured memory-spanning user facts, preferences, and working memory. By introducing a hierarchical stage-wise process reward and reward-decomposed contrastive refinement, SALIMORY provides isolated supervision for distinct memory operations (selective filtering, consolidation, and cue-driven recall) end-to-end. SALIMORY cuts memory-attributed failures by one-third, outperforms the state-of-the-art by over 10% in end-to-end accuracy, and more than doubles the Good Personalization rate.

arXiv

Holographic Invariant Storage: Design-Time Safety Contracts via Vector Symbolic Architectures

Arsenios Scrivens

We introduce Holographic Invariant Storage (HIS), a protocol that assembles known properties of bipolar Vector Symbolic Architectures into a design-time safety contract for LLM context-drift mitigation. The contract provides three closed-form guarantees evaluable before deployment: single-signal recovery fidelity converging to $1/\sqrt{2} \approx 0.707$ (regardless of noise depth or content), continuous-noise robustness $2Φ(1/σ) - 1$, and multi-signal capacity degradation $\approx\sqrt{1/(K+1)}$. These bounds, validated by Monte Carlo simulation ($n = 1{,}000$), enable a systems engineer to budget recovery fidelity and codebook capacity at design time -- a property no timer or embedding-distance metric provides. A pilot behavioral experiment (four LLMs, 2B--7B, 720 trials) confirms that safety re-injection improves adherence at the 2B scale; full results are in an appendix.

arXiv

GR-SAP: Generative Replay for Safety Alignment Preservation during Fine-Tuning

Zhouxiang Fang, Jiawei Zhou, Hanjie Chen

Recent studies show that the safety alignment of large language models (LLMs) can be easily compromised even by seemingly non-adversarial fine-tuning. To preserve safety alignment during fine-tuning, a widely used strategy is to jointly optimize safety and task objectives by mixing in the original alignment data, which is typically inaccessible even for open-weight LLMs. Inspired by generative replay in continual learning, we propose Generative Replay for Safety Alignment Preservation (GR-SAP), a unified framework that synthesizes domain-specific alignment data from LLMs and integrate them during downstream adaption to preserve safety alignment. Theoretical and empirical analyses demonstrate this synthetic data serves as a reliable proxy for the original alignment data. Experiments across various models and downstream tasks show that GR-SAP substantially mitigates fine-tuning-induced safety degradation while maintaining comparable downstream performance. Our code is available at https://github.com/chili-lab/gr-sap.

arXiv

GhazalBench: Canonical Verse Access in LLMs across Persian Ghazals and Shakespearean Sonnets

Ghazal Kalhor, Yadollah Yaghoobzadeh

Persian poetry plays an active role in Iranian cultural practice, where verses by canonical poets such as Hafez and Saadi are frequently quoted, paraphrased, or completed from incomplete cues. Supporting such interactions requires language models to reliably access canonical verses from semantic and lexical information. We introduce GhazalBench, a benchmark for evaluating how large language models (LLMs) access the canonical surface forms of Persian ghazal under usage-grounded conditions. Unlike prior work that primarily treats memorization as a liability, GhazalBench studies settings in which access to exact wording is functionally important. The benchmark evaluates completion and recognition under varied semantic and lexical cues. Across proprietary and open-weight multilingual LLMs, exact completion remains challenging, while recognition is substantially stronger. Performance also varies across poets and model families. Parallel experiments on Shakespearean sonnets yield markedly higher completion for several models, consistent with differences in exposure to canonical texts. {Additional experiments suggest that post-training may reduce direct continuation-based access to canonical text.} Our findings motivate evaluation frameworks that distinguish production from recognition and assess access to culturally significant canonical texts. GhazalBench is available at https://github.com/kalhorghazal/GhazalBench.

arXiv

Comparing Uncertainty Measurement and Mitigation Methods for Large Language Models: A Systematic Review

Toghrul Abbasli, Kentaroh Toyoda, Yuan Wang, Leon Witt, Muhammad Asif Ali, Yukai Miao, Dan Li, Qingsong Wei

Large Language Models (LLMs) have been transformative across many domains. However, hallucination, i.e., confidently outputting incorrect information, remains one of the leading challenges for LLMs. This raises the question of how to accurately assess and quantify the uncertainty of LLMs. Extensive literature on traditional models has explored Uncertainty Quantification (UQ) to measure uncertainty and employed calibration techniques to address the misalignment between uncertainty and accuracy. While some of these methods have been adapted for LLMs, the literature lacks an in-depth analysis of their effectiveness and does not offer a comprehensive benchmark to enable insightful comparison among existing solutions. In this work, we fill this gap via a systematic survey of representative prior works on UQ and calibration for LLMs and introduce a rigorous benchmark. Using three widely used reliability datasets, we empirically evaluate seven related methods, which justify the significant findings of our review. Finally, we provide outlooks for key future directions and outline open challenges. To the best of our knowledge, this survey is one of the first dedicated studies to review the calibration methods and relevant metrics for LLMs.

arXiv

The Shrinking Landscape of Linguistic Diversity in the Age of Large Language Models

Zhivar Sourati, Farzan Karimi-Malekabadi, Meltem Ozcan, Colin McDaniel, Alireza Ziabari, Jackson Trager, Ala Tak, Meng Chen, Fred Morstatter, Morteza Dehghani

Language is far more than a communication tool; it encodes a wealth of information about a person's identity, psychological state, and social context, providing valuable insights for diverse fields including psychology, marketing, and healthcare. Across three studies spanning seven datasets in different domains and over 880,000 texts, we show that the widespread adoption of large language models (LLMs) as writing assistants is linked to declines in linguistic diversity, interfering with the societal and psychological insights language provides. While core content is retained when LLMs polish and rewrite texts, LLMs also homogenize writing styles, reducing writing-complexity variance by a statistically significant 21-50% across datasets and models (p <= .05), and amplify patterns associated with dominant characteristics while suppressing others, emphasizing conformity over individuality. These trends hold across different LLMs, prompts, and contexts, with potential implications for diagnostic processes, personalization efforts, hiring assessments, and cultural preservation.

Keep browsing

More papers from top AI conferences

30+ top conferences, by conference, year, and topic.