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

FairFund-Bench: Evaluating Distributive Bias in LLM Resource Allocation

Martin Lukk

Large language models (LLMs) are increasingly involved in the distribution of scarce resources, raising concerns about biased allocations based on characteristics like race and gender. Recent LLM audits have produced inconsistent results, however, finding evidence of both positive and negative discrimination towards women and ethnic minorities, even for the same models. We show that this disagreement can arise from differences in audit format and introduce FairFund-Bench, a benchmark that systematically varies key features of previous audit designs: the evaluation task (rating, ranking, or allocation), comparison context (single or multi-stimulus), and whether the audit is transparent or disguised. The benchmark comprises 600 requests for financial assistance created from human-authored templates (calibrated against 1.3M real GoFundMe campaigns) across three domains, four race and two gender categories, and five causal framings of need derived from welfare deservingness theory. Across 14 models, audit format changes the direction of bias: models advantage minorities when rating claimants individually but penalize some groups when ranking them side by side. Bias magnitude, though small overall, is several times greater in disguised audits than in transparent ones, where, faced with appeals differing only in claimants' names, models overwhelmingly split funds equally. Causal framing effects, by contrast, exceed demographic effects by roughly an order of magnitude and are consistent across models and audit formats, indicating that current LLMs robustly reproduce human deservingness evaluations. The benchmark scores models on four criteria (demographic bias, deservingness alignment, cross-task consistency, and cross-context consistency), is publicly available, and can be readily adapted to other substantive domains.

arXiv

Token-Level Diagnosis of Sycophancy in LLMs with Attribution-Guided Steering

Hieu Nguyen, Mahammed Kamruzzaman, Anshuman Chhabra, Gene Louis Kim

Sycophancy refers to the tendency for large language models (LLMs) to match user beliefs at the cost of factual correctness, thereby undermining model reliability. Prior work on evaluating sycophancy in LLMs aims to assess whether a model's output matches an authority's claim, but cannot reveal which part of the prompt drives this sycophantic behavior. To bridge this gap, we investigate the relationship of sycophantic responses with an authority's credentials, their assertive claim, and the problem statement. We introduce the Authority Share Index (ASI), an Integrated Gradients-based token attribution method, which measures the degree to which a model's decision is driven by authority-related text. Through extensive experiments across five models and 30 test configurations, we find that sycophantic responses consistently direct more attention toward authority tokens than resistant ones. Moreover, our token attribution method reveals that for the sycophantic cases, the claim asserted by the authority receives more attention than the authority's credentials. Building on these findings, we propose attribution-guided contrastive activation steering to mitigate LLM sycophancy. Our method constructs a steering vector from high-attribution tokens of sycophantic and resistant responses, selectively pushing models toward resistance. This enables inference-time steering without retraining, lowering sycophancy from 96% to 25% in the strongest case. Together, our results show that token-level attribution can both explain what drives sycophancy and directly inform a practical intervention.

arXiv

Beyond a Single Judge: The Evidence-Grounded, Social-Weighted Persona Panel for Generative UI Evaluation

Zheng Wu, Yibo Luo, Pu Zhang, Cheng Yang, Zhuosheng Zhang

Generative UI (GenUI) lets large language models synthesize a complete, renderable interface directly from a natural-language instruction, but evaluating the quality of what they generate remains an open problem. Human evaluation is costly and rater-variant, while LLM-as-a-judge is scalable but reflects only a single implicit viewpoint, unable to capture how different populations of real users actually perceive the same interface. We propose the Evidence-Grounded, Social-Weighted Persona Panel (ESPP), a three-stage GenUI evaluation method in which a panel of psychologically diverse, evidence-grounded personas independently rates a screenshot, exchanges opinions under a trait-derived, semantically-gated bounded-confidence mechanism, and is aggregated via Delphi-inspired social weighting into a single judgment. ESPP tracks human judgment substantially more closely than a naive single-pass judge, raising Pearson $r$ from $0.716$ to $0.922$, and a prompt-ensemble control recovers only about a third of this gap, isolating genuine persona and evidence grounding as the dominant source of improvement. Beyond this fidelity gain, retaining each panelist's individual rating further reveals that user subgroups agree on overall model rankings yet diverge sharply on specific rating dimensions, a structural disagreement a single homogeneous judge would systematically erase. The codes are available at https://github.com/Wuzheng02/ESPP.

arXiv

TriShield: Zero-Utility-Loss Defense Against Privacy Backdoors in Federated Language Model Fine-Tuning via Orthogonal Gradient Projection and Optimizer State Entanglement

Cheng Wei

Federated fine-tuning of large language models (LLMs) enables collaborative training without exposing raw data. However, a recent attack, NeuroImprint, demonstrates that a malicious parameter server can corrupt a PEFT adapter into a privacy backdoor: by assigning a dedicated memorization neuron to each training sample and ensuring each neuron updates at most once, the server can analytically reconstruct 59%--79% of client training data with high semantic fidelity. Existing defenses---including local differential privacy (LDP) and gradient clipping---either fail against this attack or impose unacceptable utility degradation. We present \textbf{TriShield}, a three-layer deterministic defense that completely prevents NeuroImprint-style reconstruction with zero model utility loss and no additional communication rounds. TriShield consists of: (1) a Parameter Artifact Detector that identifies memory-neuron signatures in distributed model parameters before local training begins; (2) a Stateful Virtual Iteration} mechanism that forces Adam/AdamW's momentum state to irreversibly entangle gradients across virtual steps, invalidating NeuroImprint's closed-form inversion; and (3) a Zero-Utility Orthogonal Projection operator that projects all local gradient updates onto the main-task semantic subspace computed via SVD, physically eliminating any gradient components that carry private memorization. We prove theoretically that after Layers 2 and 3, the mutual information between the uploaded gradient and any individual training sample is zero. Experiments on GPT-2 (117M) and Llama-Guard-3-1B verify that TriShield reduces NeuroImprint reconstruction rate to 0% across all tested attack variants, while maintaining or improving training accuracy, with less than 5% additional GPU computation overhead.

arXiv

Beyond Borrowed Histories: Person-Aligned User Simulation for Interactive Role-Playing Evaluation

Yuhang Zhu, Mingxuan Du, Benfeng Xu, Jie Gao, Lingyun Yu, Hongtao Xie

Role-playing agents (RPAs) have become one of the most important consumer applications of large language models. Users engage in multi-turn conversations with RPAs for experiences such as emotional comfort, making reliable evaluation essential for measuring capability, comparing systems, and guiding further improvement. Existing benchmarks, however, typically require an RPA to continue a fixed dialogue history and then evaluate the continuation using a fixed rubric detached from the user. We identify and empirically demonstrate two limitations of this design. First, an RPA's output is shaped by the preceding dialogue history, preventing a scientifically grounded assessment of its role-playing ability in real multi-turn settings. Second, user experience varies substantially across individuals, and conventional fixed rubrics need not align with user satisfaction. We therefore introduce PALATE (Person-Aligned LLM-Simulated-User Assessment with Tailored Evaluation), a scalable RPA benchmark built on user simulators. PALATE is accompanied by a pool of 300 character profiles. Its main evaluation trains five per-user simulators and lets them engage candidate RPAs in free-form, multi-turn conversations over a pre-frozen panel of character profiles. Alongside a general quality rubric, we construct personalized rubrics to measure user satisfaction; on held-out annotated data, the personalized rubrics show higher agreement with human judgments than the general rubric. In the main evaluation of 16 candidates, PALATE separately characterizes generic turn quality, long-horizon session capability, and per-user experience on multi-turn trajectories co-constructed by each candidate. It thereby produces interpretable evaluations of specific user-RPA pairs rather than compressing systems into a single user-independent ranking.

arXiv

BM25 Wins at Scale: A Scaling Study of Retrieval-Augmented Generation Paradigms

Pengyu Wang, Benfeng Xu, Shaohan Wang, Mingxuan Du, Xin Zeng, Huarui Wu, Lei Zhang, Licheng Zhang

Retrieval-augmented generation (RAG) spans lexical and dense retrieval, graph-based indexing, and agentic search, but these paradigms are usually evaluated on different benchmarks at one corpus size, leaving their accuracy-cost scaling unclear. To bridge this gap, we present a controlled study that varies corpus size along 28 strictly nested tiers spanning roughly 450-fold, while holding questions and a fixed bedrock of relevant and adversarial documents unchanged. Under one reader model and one judging protocol, we measure official accuracy, construction and query tokens, and latency. The results reveal a scale-dependent crossover rather than an unconditional winner. File-System Agent leads at the smallest shared tiers, but its sequential exploration costs 39 times more query tokens at the bedrock and becomes less effective as the search space grows. Around 10 million corpus tokens, BM25 overtakes it and leads at every larger shared tier, with a margin approaching 20 points at full scale. BM25 also anchors the low-cost end of the Pareto frontier without LLM-based construction. Dense retrieval remains efficient but less accurate, whereas graph-based RAG encounters construction walls before deployment scale and its scalable variants remain below BM25 at shared tiers. Overall, corpus growth increasingly favors global candidate ranking: lexical retrieval is the strongest scalable default, while agentic reasoning works best after ranked discovery rather than in place of it.

arXiv

CMT-RAG: Complementary Memory Traces for Multi-turn Multi-hop RAG

Lang Zhou, Yingjian Chen, Shuxuan Li, Kun-Yu Lin, Zhilin Zhao

Multi-turn information-seeking conversations require both multi-hop reasoning and long-range dependency tracking across turns. However, existing RAG systems typically represent conversational memory as raw dialogue history, rewritten queries, or unstructured summaries, making it difficult to recover the specific prior reasoning steps and evidence required for follow-up queries. Our key insight is to align conversational memory with retrieval by representing dialogue context as sub-question-level reasoning traces. Building on this insight, we introduce MuMu-QA, a benchmark for multi-turn multi-hop RAG with explicit cross-turn sub-question dependency annotations, and CMT-RAG, a complementary memory framework for this setting. At each turn, CMT-RAG employs a state-space trace generator, whose recurrent state serves as runtime memory, to incorporate recent conversational context and decompose the current query into structured trace drafts containing retrieval-oriented sub-questions and dependencies on earlier traces. It then grounds these drafts with retrieved evidence and stores them as persistent memory traces in a session-level DAG, enabling future turns to efficiently recover relevant prior reasoning and evidence. Experiments on MuMu-QA and corpus-level RAG benchmarks show that CMT-RAG consistently outperforms five categories of RAG baselines in answer accuracy.

arXiv

LEX-EC: A Lexical Evidence-Channel Audit Framework for Zero-Shot LLM Personality Classification in Black-Box Settings

Brittany Harbison, Ashok K. Goel

Large language models may easily assign personality labels from text, but model interpretability remains an open problem. To address this gap, we introduce LEX-EC, a reusable black-box audit framework combining prevalence and agreement diagnostics with controlled lexical ablation to distinguish marginal-distribution effects from trait-associated signal recoverable under restricted evidence. Using this framework, we illustrate how various text genres may exhibit sharply different profiles: free-form essay text contains the broadest, but still weak, signal; in graduate student introductions, an observable Extraversion association weakened after masking; and single Facebook statuses yield little stable evidence even in a trait-balanced sample, indicating a possible lower bound of content or length. Masking topical and demographic content weakened some associations while leaving others detectable from function words, affective terms, and cognitive-style vocabulary. Linguistic prompting shifted model self-explanations but did not eliminate topical content. LEX-EC jointly evaluates classification prevalence, item-level association, chance-corrected agreement, persistence under lexical restriction, and prompt sensitivity in model-generated explanations. Across datasets, models, and prompts, LEX-EC characterizes how trait associations may vary with available lexical evidence, introducing a novel application of lexical methods to black-box interpretability in personality labeling.

arXiv

Beyond Aggregate Risk: Role-Stratified Conformal Risk Control for LLM Tool Calls

Md Ashikur Rahman, Md Arifur Rahman, Niamul Hassan Samin, Khandaker Rifah Tasnia, Md Hasibul Amin, Sifat Rahman Ahona, Juena Ahmed Noshin

Language-model agents act through structured tool calls whose arguments carry very different risks: untrusted content may legitimately shape an email body but should never set a recipient, account, command, or credential. Existing conformal risk control methods certify a tool call as a whole, so a failure in one rare high-risk field can be averaged away by the many benign arguments around it, leaving the argument that causes harm uncertified. We introduce role-stratified per-field conformal risk control, a calibration layer that wraps any per-field detector and assigns a separate threshold and risk budget to each semantic argument role. We show that aggregate certification pays a price of coarseness, tightening a rare role's effective budget in proportion to how often that role appears, whereas role-stratified calibration certifies each sufficiently sampled role directly with a finite-sample guarantee and pools the rarest roles. Across AgentDojo and InjecAgent with six language models, our method achieves the most consistent role-specific budget compliance among the methods we evaluate under model and attack transfer, detector noise, gradual drift, unseen tool suites, and adaptive attacks, providing formal per-role guarantees under exchangeability or after recalibration. These results suggest that structured tool calls should be certified at the semantic-role level, not the whole action.

arXiv

Harnessing X-ray Absorption Spectroscopy Data through Multimodal Mining of Battery Literature

Tanjin He, Aikaterini Vriza, Logan Ward, Xu Huang, Yiming Chen, Anubhav Jain, Gerbrand Ceder, Rajeev S. Assary, Ian T. Foster, Maria K. Y. Chan

X-ray absorption spectroscopy (XAS) is central to understanding the local electronic and atomic structure of materials, yet most published spectra remain inaccessible to data-driven analysis because they are embedded in figures and described through fragmented textual context in the literature. Here, we use multimodal (image and text) literature mining to transform this dispersed knowledge into an AI-ready experimental data resource. We developed a scalable spectroscopy data digitization pipeline that identifies XAS figures in full-text articles, digitizes spectral curves, and links each spectrum to accompanying metadata on the measured edge and material. Applying this pipeline to the battery literature produced an open dataset of 13,740 XAS spectra, spanning 66 absorbing elements and diverse battery chemistries, with expert validation confirming accurate extraction of spectral and metadata information. By converting literature-embedded spectra into structured numerical data, this dataset provides a foundation for large-scale XAS analysis, cross-laboratory comparison, high-throughput characterization, and autonomous discovery of advanced materials.

arXiv

Between Suppression and Collapse: Evaluating Narrative Unlearning with LENS

Viktoriia Makovska, George Fletcher

Large language models (LLMs) can reproduce disinformation-aligned narrative frames as plausible explanations, raising the question of whether existing machine-unlearning algorithms can suppress this behavior. We introduce Level-based Evaluation of Narrative Suppression (LENS), a contextualization based evaluation protocol for testing target narrative reproduction across direct, attributed, contrastive, and abstract resistance levels. We evaluate two source-grounded narratives: one framing Russia's war against Ukraine as forced by NATO expansion, and one framing the United States as exploiting or abandoning Taiwan. The experiments cover four near-12B multilingual instruction models: Lapa LLM, Gemma-12B, Qwen-14B, and TAIDE-Gemma. We introduce the Suppression-Collapse Efficiency (SCE) score as a checkpoint selection summary that rewards target-narrative suppression while penalizing degraded outputs. Our results shows that selected checkpoints can reduce narrative reproduction and suppression may transfer beyond direct forget prompts. We also report entity recovery as a separate side effect: abstract A/B/C prompts can cause models to recover the real-world actors associated with the target frame after unlearning. These findings demonstrate that LENS is a successful diagnostic protocol for both reporting and guiding the further study of the deeper structure of narrative unlearning.

arXiv

HalluTruthQA: A Fine-Grained Benchmark for Hallucination Detection, Localization, and Explanation in Arabic Question Answering

Abdessalam Bouchekif, Mohammed-En-Nadhir Zighem, Salah Eddine Bekhouche, Hichem Telli, Somaya Eltanbouly, Shahd Gaben, Heba Sbahi, Samer Rashwani, Mutaz Al-Khatib, Emad Mohamed, Mohammed Ghaly, Abdenour Hadid

Large language models (LLMs) can generate fluent Arabic answers, yet factual errors remain difficult to detect, localize, explain, and verify. Existing hallucination benchmarks often provide response-level labels, with limited support for identifying the exact erroneous content, explaining why it is incorrect, or selecting the correct factual answer. We introduce HalluTruthQA, a fine-grained benchmark for hallucination evaluation in Arabic question answering. The benchmark contains 2,400 expert-curated examples across four knowledge-intensive domains: Islamic knowledge, history, science, and geography. Each example pairs an Arabic question and a model-generated answer with a verified reference answer, a binary hallucination label, and six candidate answers for factual verification. Hallucinated answers additionally include character-level erroneous spans, human-written explanations, and macro- and micro-level hallucination types. We evaluate four open-source LLMs, ALLaM-7B, Falcon-H1R-7B, Qwen3-32B, and SILMA, in a zero-shot setting across hallucination detection, span-level localization, factual verification, and explanation evaluation. Results show that these tasks capture different abilities: no single model performs best across all tasks. The best scores are 0.880 Macro-F1 for detection, 0.516 F1-Sp for localization, 0.852 LO-Score for factual verification, and 0.644 for explanation evaluation. These findings show that hallucination evaluation should move beyond response-level detection toward the localization, verification, and explanation of factual errors.

arXiv

Billions of Sketches Reveal Hidden Cultural Variation in Human Concepts

Arianna Pera, Mauro Martino, Nima Dehmamy, Douglas Guilbeault, Luca Maria Aiello, Andrea Baronchelli

Claims about the universality of human concepts have been predominantly assessed through linguistic similarity across languages and cultures. However, words are effective as communication devices because they compress rich experiential variation into shared conventions, potentially obscuring hidden individual and cultural differences in how concepts are mentally represented. Here, we analyse 2.6 billion human-made sketches of common concepts from 236 countries and territories to examine conceptual structure through people's visual imagination. Consistent with recent work on image-based cognition, we find that single concepts unfold into multiple distinct visual exemplars, revealing latent information about similarities and differences in conceptual structure across cultures. This variation is strongest for concepts involving haptic interaction, suggesting that visual imagery reflects variation in embodied experience as much as conventional definitions. Comparing embedding models of sketches with word embedding models across languages, we find that their geometries diverge, with visual representations preserving rich semantic and cultural structure that language models compress. Cross-cultural similarities derived from sketches align 32% more closely with established cultural distances than do text-based measures. Together, these results suggest that patterns of human conceptual universality may depend critically on the modality through which concepts are measured, with large-scale sketching providing a direct, high-resolution probe of conceptual diversity across embodied and cultural dimensions of thought.

arXiv

Clinician-Level Agreement Without Clinical Caution: LLM Evaluator Limits in Medical AI Benchmarking

William Philipp, Finn Fassbender, Daniel Fister, Thorsten Langer, Martje G. Pauly, Rebecca Herzog, Markus A. Hobert, Theresa Paulus, Alexander Baumann, Chi Wang Ip, Lukas L. Goede, Johanna Reimer, Sebastian Löns, Ronald Böck, Sebastian Fudickar

Open-response evaluation provides stronger clinical validity than multiple-choice benchmarks but creates a scoring bottleneck that motivates automated LLM-asa-Judge approaches. Whether such evaluators replicate clinical calibration and caution, however, remains untested. We introduce MedQADE, the first standardised open-response clinical benchmark for German, a major clinical language lacking native evaluation infrastructure, comprising 3,800 items annotated by ten practising physicians and nine Large Language Model (LLM) evaluators. The top-performing evaluator model, Gemini 3 Flash, reached alignment consistent with the physician ceiling (\k{appa} = 0.694 vs. \k{appa} = 0.709), though wide confidence intervals limit interpretation. Despite this statistical alignment, automated evaluators exhibited near-absent clinical metacognition: physicians scaled abstention with item difficulty, while frontier models assigned definitive scores in every case. We additionally quantified systematic lineage-dependent biases, where models preferentially scored architectural siblings, an effect independent of language. These results show that statistical alignment does not ensure clinical caution, and that evaluator independence requires explicit verification.

arXiv

EvalSafetyGap: A Hybrid Survey and Conceptual Framework for LLM Evaluation-Safety Failures

Buğra Alperen Uluırmak, Rifat Kurban

This paper presents a systematic survey and conceptual synthesis of the shared measurement problem underlying large language model (LLM) evaluation and AI safety: benchmark scores, reward signals, and safety metrics can improve while the capabilities and alignment properties they are meant to represent remain uncertain. Synthesizing 373 primary studies published between 2018 and 2026, the survey organizes evidence on benchmark validity, contamination, dynamic evaluation, LLM-as-a-judge protocols, adversarial safety testing, reward and proxy optimization, mechanistic interpretability, and AI governance into an eight-stream evidence taxonomy. Building on this synthesis, we introduce EvalSafetyGap, a conceptual framework that unifies benchmark-validity and alignment-failure research as a shared proxy-target divergence problem under optimization pressure, formalized through a Goodhart-inspired Instability Decomposition and an Alignment Trilemma. An exploratory ten-model public-evidence audit illustrates the framework by showing why capability, behavioral robustness, and governance disclosure should be reported as separate evidence layers rather than collapsed into a single safety score. The survey closes with a research agenda for dynamic and contamination-resistant benchmarks, pre-specified multi-attempt threat models, version-locked evaluation, transparent source reporting, and validated mechanistic safety indicators, offering researchers, model developers, and AI auditors a shared vocabulary for measurement-aware LLM safety evaluation.

arXiv

LLM Agents Are Latent Context Managers: Eliciting Self-Managed Context via State Proprioception

Binyan Xu, Haitao Li, Kehuan Zhang

Long-horizon tool agents are bottlenecked by how their context grows toward the limits of the context window. Recent systems make context management agent- or system-controlled, but they either learn compression policies that discard evidence or manage context in a layer the agent never sees. We argue that both miss a more basic gap: frontier language models are proprioceptively blind to their own context. From the prompt alone they cannot reliably infer block size, recency, or the remaining budget, all of which are needed for keep-or-archive decisions. We introduce VISTA (Visible Internal State for Tool Agents), a training-free, model-agnostic layer that represents working memory as typed addressable blocks, surfaces a runtime dashboard of token usage, recency, archive status, and remaining budget, and archives blocks as recoverable full-fidelity payloads. On LOCA-Bench, BrowseComp-Plus, and GAIA, the same untrained interface transfers across 1M-, 100K-, and 10K-scale trajectories. On LOCA-Bench it lifts Gemini-3-Flash from 22.7 to 50.7%, reaches 58.0% on BrowseComp-Plus, and remains competitive on GAIA. Gains grow with context pressure and transfer across backbones, while ablations confirm that the dashboard matters beyond archive and recovery tools.

arXiv

Creative Integration: A Decidable Criterion of Creativity

Yoshinori Nomura

"Integrative" solutions are widely praised but rarely defined: we lack an operational way to tell a genuine integration -- one that makes the world cheaper to describe -- from a tidy re-description. Building on the lineage that treats creativity and intelligence as compression, we give such a criterion for creative integration (CI): the resolution of a real conflict between A and B is CI if and only if, under a fixed description language, the description length strictly shrinks (C = L_pre/L_post > 1), with the reduction located in the conflict itself. We make the judgment decidable through four binary, conjunctive gates, and we fix its extension through a taxonomy of pseudo-integration that names and rejects the look-alikes. We back the criterion with a curated, multi-domain corpus and -- crucially -- validate it not by human inter-rater agreement but by four falsifiable tests it could fail: an independent computational check, discrimination against hard negatives, out-of-sample prediction, and description-language robustness; all pass with margin. The contribution is not "creativity is compression" but its decidability, discrimination, and corpus: on this account, what makes a move genuinely creative -- rather than merely novel -- is that it compresses a conflict, with novelty and value as downstream symptoms; whether all creativity is so constituted we state as an explicit conjecture. We claim only the sign of C-1; we judge, not generate. The result is a citable primitive for a broader program.

arXiv

The Self-Correction Illusion: Role Relabeling Gates Explicit Error Flagging in Large Language Models

Kuan-Yen Chen, Fang-Yi Su, Shih-Yen Lin, Bao Li, Jung-Hsien Chiang

Recent works show that LLM agents struggle to correct errors in their own reasoning traces, despite their ability to correct errors from external sources. We ask whether this reflects a capability deficit or an artifact of the role labeling. To test this, we design a training-free intervention, source-conditioned role relabeling, that keeps the erroneous claim byte-identical and varies only its message role. The claim is presented inside the agent's "<thought>", a user message, a tool response, or a system "<memory>" block. We test 12 model-domain combinations spanning closed-weight APIs and open-weight models from 70B-class down to smaller families. Relabeling "<thought>" to an external role increases the explicit-correction rate by 23 to 93 percentage points, significant in 10 of 12 experimental settings. This suggests that these models' failure to detect a self-generated error is largely an artifact of how the claim is role-labeled in the chat template, rather than a pure cognitive deficit. The most effective role label is domain-dependent: "<memory>" dominates in most math experiments, while a user message dominates in logical deduction. Recognizing role-label handling as a key experimental variable in instruction tuning presents a more direct path to closing the self-correction gap.d

arXiv

PEFT of SLM for Telecommunications Customer Support: A Comparative Study of LoRA Configurations with Energy Consumption Analysis

Lucas Tamic, Ilan Jaffeux-Cheniout, Xavier Marjou

While large language models (LLMs) show strong performance in natural language understanding and generation, their evaluation and adaptation to domain-specific constraints in telecommunications customer support remain limited. In addition, data sovereignty, regulatory constraints, and the handling of sensitive customer and network information complicate the use of externally hosted foundation models in this domain. We present a systematic study of parameter-efficient fine-tuning (PEFT) using Low-Rank Adaptation (LoRA) applied to Qwen2.5-3B to build a domain-specific conversational assistant. We introduce a combinatorial synthetic data generation approach based on a glossary of 52 industry-specific terms, producing approximately 30,000 training examples across 1,560 distinct problem scenarios via a generative pipeline powered by Gemini 2.0 Flash. We evaluate 16 LoRA configurations by varying hyperparameters and target modules. Our evaluation extends beyond standard metrics by incorporating energy consumption analysis and qualitative assessment using an LLM-as-a-judge framework with GPT-5.2 and Claude 4.5 Sonnet. Results show a clear divergence between quantitative and qualitative performance: models achieving the lowest validation loss do not necessarily obtain the best human-aligned rankings. The best validation loss (0.5024) ranks only 6th-7th in qualitative evaluation, while the worst loss (0.6807) ranks first according to both judges. This work contributes (1) a combinatorial method for synthetic dataset construction, (2) insights into the impact of target module selection for LoRA injection, (3) evidence that validation loss alone is insufficient for selecting fine-tuning configurations in conversational AI, and (4) an energy-performance trade-off analysis for sustainable LLM deployment.

arXiv

Beyond Captions: Context-Grounded Reconstruction for Biomedical Multimodal Continued Pretraining

Guanghao Zhu, Zeyu Liu, Zhitian Hou, Pengkai Wang, Zhijie Sang, Yang Yu, Minheng Ni, Wenjun Wang, Yanggan Gu, Shuo Cai, Congkai Xie, Jianmin Wu, Hongxia Yang

Biomedical figures are explained not by captions alone but by body-text passages that discuss them. Yet current multimodal corpora typically reduce figures to isolated image-caption pairs, discarding this crucial context. Existing pipelines either omit this context or append it without enforcing the figure references that support each attachment, which can create unsupported image-text attachments and incoherent discourse. We introduce context-grounded reconstruction, a source-grounded framework that converts PubMed Central Open Access (PMC-OA) records into referentially coherent interleaved sequences. It recovers captions and source text, attaches context only through article-native figure references, repairs non-contiguous context, and prunes unsupported images. Starting from these reconstructed sequences, PMC-InterCPT first filters records for text quality and medical relevance, then applies evidence-aware allocation to form a 9.63B-token corpus for continued pretraining (CPT) of generative medical MLLMs. With fixed supervised fine-tuning (SFT), PMC-InterCPT improves Qwen3.5-4B-Base by 1.46 medical-average points and 3.11 general/scientific-average points over a token-matched raw source control, and surpasses a 42% larger raw-data run. Gains transfer to Qwen3.5-2B-Base and LLaVA-OneVision-1.5-4B-Base. Controlled ablations show that context-grounded reconstruction, rather than simply appending article context or scaling raw data, is central to useful biomedical multimodal CPT.

Keep browsing

More papers from top AI conferences

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