{"total":24,"items":[{"citing_arxiv_id":"2606.01074","ref_index":1,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"When Is 0.1% Enough? Analyzing the Combined Effects of Dimensionality Reduction and Quantization on Text Embedding Compression","primary_cat":"cs.CL","submitted_at":"2026-05-31T07:37:34+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"Combining dimensionality reduction and quantization compresses text embeddings to 0.1% size with minimal performance loss on MTEB tasks, outperforming either technique alone.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.24307","ref_index":21,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Modernizing User Privacy Preference Measurement through GPPI: A GDPR-aligned Privacy Preference Item Bank","primary_cat":"cs.HC","submitted_at":"2026-05-23T00:36:32+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"A 527-item GDPR-aligned privacy preference item bank was developed by extracting 669 statements from 99 GDPR articles and validating them through multi-round expert consensus and semantic clustering.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.12813","ref_index":148,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"REALISTA: Realistic Latent Adversarial Attacks that Elicit LLM Hallucinations","primary_cat":"cs.CL","submitted_at":"2026-05-12T23:13:50+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"REALISTA generates semantically coherent adversarial prompts via latent-space optimization over input-dependent editing directions, achieving stronger hallucination elicitation than prior realistic attacks on open-source and reasoning LLMs.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.10606","ref_index":27,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Measuring Embedding Sensitivity to Authorial Style in French: Comparing Literary Texts with Language Model Rewritings","primary_cat":"cs.CL","submitted_at":"2026-05-11T14:05:46+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"Embeddings reliably capture authorial stylistic features in French literary texts, and these signals persist after LLM rewriting while showing model-specific patterns.","context_count":1,"top_context_role":"background","top_context_polarity":"background","context_text":"ing the Classical in Marguerite Yourcenar, volume 271 ofFaux Titre. Brill / Rodopi, Leiden. Corinna Cortes and Vladimir Vapnik. 1995. Support- vector networks.Machine Learning, 20(3):273-297. Louis-Ferdinand Céline. 1932.Voyage au bout de la nuit. Denoël et Steele. Antoine Silvestre de Sacy. 2025. Hypersegmentation du discours chez louis-ferdinand céline.Corpus, (27). Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. Bert: Pre-training of deep bidirectional transformers for language understand- ing.arXiv preprint arXiv:1810.04805. Géraud Faye, Benjamin Icard, Morgane Casanova, Ju- lien Chanson, François Maine, François Bancilhon, Guillaume Gadek, Guillaume Gravier, and Paul Égré. 2024. Exposing propaganda: an analysis of styl-"},{"citing_arxiv_id":"2605.10430","ref_index":29,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Real vs. Semi-Simulated: Rethinking Evaluation for Treatment Effect Estimation","primary_cat":"cs.LG","submitted_at":"2026-05-11T12:04:02+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Large-scale study finds that counterfactual metrics on semi-simulated data do not select the same estimators as observable metrics on real data, and benchmark rankings fail to transfer.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.10021","ref_index":6,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Enhancing Healthcare Search Intent Recognition with Query Representation Learning and Session Context","primary_cat":"cs.IR","submitted_at":"2026-05-11T05:44:54+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"Clustering-based query representations with a novel multi-intent loss and a concordance rate metric improve healthcare search intent classification on two real-world log datasets.","context_count":1,"top_context_role":"method","top_context_polarity":"use_method","context_text":"that evaluates the accuracy of predicted labels for each query. This process typically involves techniques like Binary Cross-Entropy (BCE) applied across all 𝐼 labels. Labels are then selected based on a thresholding mechanism to form the setY 𝑞. 3.2 Proposed Approach RQ1: Enhancing Query Representation LearningAs shown in Figure 2a, our methodology utilizes the transformer-based query encoder (i.e. BERT [6]) as the initial query encoder. We enhance query representations through contrastive loss functions L𝑐𝑜𝑛𝑡𝑟𝑎𝑠𝑡 , tailoring the encoder to the intents of health search queries. The con- trastive loss functions distinguish between pairs of queries clicked on the same document denoted as (𝑞, 𝑞∗) and those clicked on dif- ferent documents denoted as (𝑞, 𝑞−), ensuring the encoder captures"},{"citing_arxiv_id":"2605.04295","ref_index":3,"ref_count":2,"confidence":0.98,"is_internal_anchor":true,"paper_title":"LLMs Uncertainty Quantification via Adaptive Conformal Semantic Entropy","primary_cat":"cs.LG","submitted_at":"2026-05-05T20:56:11+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"ACSE estimates LLM uncertainty via adaptive semantic entropy clustering with conformal prediction guarantees, reporting higher AUROC than token entropy baselines on datasets like TriviaQA.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.00618","ref_index":16,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Is Textual Similarity Invariant under Machine Translation? Evidence Based on the Political Manifesto Corpus","primary_cat":"cs.CL","submitted_at":"2026-05-01T12:41:00+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Machine translation preserves embedding similarity structure for ten languages but distorts it for four in the Manifesto Corpus, via a new non-inferiority testing framework.","context_count":1,"top_context_role":"method","top_context_polarity":"use_method","context_text":"the basis of manifesto fragments. Our label data source is the Man- ifesto Research Project Dataset [ 35], a companion dataset to the Manifesto Corpus, where each party is assigned to one of 12 party families. We use a random forest classifier [10]. To evaluate classifier performance, we use macroF1 scores [48] and Matthews Correlation Coefficient [ 16]. We compare distribu- tions of both metrics for original-language, multilingual, and post- translation models. To measure agreement between classifiers run under different embedding models and text versions, we use the ad- justed Rand index [52, 26], for which we run the same statistical tests as for correlations between similarity matrices. Clustering"},{"citing_arxiv_id":"2604.27641","ref_index":12,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Semantics-Aware Hierarchical Token Communication: Clustering, Bit Mapping, and Power Allocation","primary_cat":"eess.SP","submitted_at":"2026-04-30T09:34:05+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"H-TokCom groups tokens by semantic similarity and protects cluster-level bits with higher power, raising semantic similarity from 0.206 to 0.279 at 3 dB SNR on COCO data.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.17257","ref_index":32,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"REZE: Representation Regularization for Domain-adaptive Text Embedding Pre-finetuning","primary_cat":"cs.CL","submitted_at":"2026-04-19T04:41:55+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"REZE controls representation shifts in contrastive pre-finetuning of text embeddings via eigenspace decomposition of anchor-positive pairs and adaptive soft-shrinkage on task-variant directions.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.16770","ref_index":31,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Exploring Ethical Concerns of Mobile Applications from App Reviews: A Literature Survey","primary_cat":"cs.SE","submitted_at":"2026-04-18T01:30:06+00:00","verdict":"ACCEPT","verdict_confidence":"LOW","novelty_score":4.0,"formal_verification":"none","one_line_summary":"A systematic literature survey of 37 studies synthesizes methods for identifying ethical concerns from app reviews and proposes a research agenda focused on automated detection.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.07553","ref_index":1,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"TR-EduVSum: A Turkish-Focused Dataset and Consensus Framework for Educational Video Summarization","primary_cat":"cs.CL","submitted_at":"2026-04-08T19:53:39+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Presents TR-EduVSum dataset and AutoMUP consensus framework for generating gold-standard summaries from multiple human annotations of Turkish educational videos.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2508.20086","ref_index":12,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Detecting Malicious Intents in Smart Contracts with Pre-trained Programming Language Models","primary_cat":"cs.SE","submitted_at":"2025-08-27T17:54:15+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"SmartIntentV2 uses a pre-trained BERT model on smart contracts to achieve an F1 score of 0.9279 for detecting malicious intents, outperforming previous models and GPT-4.1.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2507.01925","ref_index":45,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"A Survey on Vision-Language-Action Models: An Action Tokenization Perspective","primary_cat":"cs.RO","submitted_at":"2025-07-02T17:34:52+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"The survey frames VLA models as pipelines that generate progressively grounded action tokens and classifies those tokens into eight types to guide future development.","context_count":1,"top_context_role":"background","top_context_polarity":"background","context_text":"To scale up data volume, LangLfP combines 10M goal-image-conditioned state-action pairs with 10K human-labeled language-conditioned samples. BC- Z [239] is one of the first works to collect a large dataset (26K robot data and 19K human videos) to study how data scaling helps generalizable policy training. It utilizes ResNet [80] and multilingual sentence encoder [45] but improves the fusion process by using multi-stage FiLM conditioning [262], which dynamically modulates visual features based on language inputs. This approach allows more fine-grained instructions grounding and decodes actions with a simpler MLP. 10.2. Transformer-Based Generalists Building on the success of scaling laws in LLMs, subsequent works have taken further steps to construct"},{"citing_arxiv_id":"2504.18902","ref_index":24,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Transformer-Empowered Actor-Critic Reinforcement Learning for Sequence-Aware Service Function Chain Partitioning","primary_cat":"cs.NI","submitted_at":"2025-04-26T12:18:57+00:00","verdict":"UNVERDICTED","verdict_confidence":"UNKNOWN","novelty_score":6.0,"formal_verification":"none","one_line_summary":"A Transformer-empowered actor-critic RL framework with ε-LoPe exploration and asymptotic return normalization improves long-term service acceptance, resource utilization, and scalability for sequence-aware SFC partitioning in heterogeneous networks.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2311.01378","ref_index":6,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Vision-Language Foundation Models as Effective Robot Imitators","primary_cat":"cs.RO","submitted_at":"2023-11-02T16:34:33+00:00","verdict":"CONDITIONAL","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"RoboFlamingo adapts open-source vision-language models for robot manipulation tasks via single-step comprehension plus an explicit policy head, outperforming prior methods on benchmarks with only light fine-tuning.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2304.10726","ref_index":16,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Usenix'23 Extended Version: Smart Learning to Find Dumb Contracts","primary_cat":"cs.CR","submitted_at":"2023-04-21T03:45:10+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"DLVA trains neural networks on bytecode to match Slither source labels at 92.7% accuracy and 0.2 seconds per contract while outperforming nine other tools at 99.7% average accuracy.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2204.01691","ref_index":15,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Do As I Can, Not As I Say: Grounding Language in Robotic Affordances","primary_cat":"cs.RO","submitted_at":"2022-04-04T17:57:11+00:00","verdict":"ACCEPT","verdict_confidence":"MODERATE","novelty_score":7.0,"formal_verification":"none","one_line_summary":"SayCan combines an LLM's high-level semantic knowledge with robot skill value functions to select only feasible actions, enabling completion of abstract natural-language instructions on a real mobile manipulator.","context_count":1,"top_context_role":"method","top_context_polarity":"use_method","context_text":"and 2 focus on the LLM and VFS components. that this description only corresponds to low level skills - it is still the role of the LLM in SayCan to interpret the high-level instruction and break it up into individual low level skill descriptions. To condition the policies on language, we utilize a pre-trained large sentence encoder language model [15]. We freeze the language model parameters during training and use the embeddings generated by passing in text descriptions of each skill. These text embeddings are used as the input to the policy and value function that specify which skill should be performed (see the details of the architectures used in the Appendix C.1). Since the language model used to generate the text"},{"citing_arxiv_id":"2104.05565","ref_index":13,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Survey on reinforcement learning for language processing","primary_cat":"cs.CL","submitted_at":"2021-04-12T15:33:11+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":2.0,"formal_verification":"none","one_line_summary":"This survey reviews reinforcement learning applications to natural language processing problems, especially conversational systems, including problem descriptions, suitability of RL, advantages, limitations, and promising directions.","context_count":1,"top_context_role":"background","top_context_polarity":"background","context_text":"signals and it can also generate more diversified texts. With this approach, the reward and the policy functions are learned alternately, following an adversarial model strategy. According to the authors, this model can generate texts with higher quality than previous proposed methods based also on GANs, such as SeqGAN [145], RankGAN [80], MaliGAN [13] and LeakGAN [43]. The adversarial text generation model uses a discriminator and a generator. The discriminator judges whether a text is real or not, meanwhile the generator learns to generate texts by maximizing a reward feedback provided by the discriminator through the use of reinforcement learning. The generation of entire text sequences that these adversarial models can accomplish helps to avoid the"},{"citing_arxiv_id":"1908.10084","ref_index":8,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks","primary_cat":"cs.CL","submitted_at":"2019-08-27T08:50:17+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":8.0,"formal_verification":"none","one_line_summary":"Sentence-BERT adapts BERT with siamese and triplet networks to produce sentence embeddings for efficient cosine-similarity comparisons, cutting computation time from hours to seconds on similarity search while matching BERT accuracy.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"1907.10710","ref_index":7,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Generic Intent Representation in Web Search","primary_cat":"cs.IR","submitted_at":"2019-07-24T20:40:19+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"GEN Encoder learns query intent embeddings from click logs as weak supervision and multi-task paraphrase training, outperforming prior methods on intent similarity and using nearest-neighbor search to cover half of unseen queries.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"1907.07366","ref_index":3,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Leveraging Linguistic Characteristics for Bipolar Disorder Recognition with Gender Differences","primary_cat":"cs.IR","submitted_at":"2019-07-17T07:37:13+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Gender-enriched syntactic pattern features from Twitter data recognize bipolar disorder with F1 scores above 91%, outperforming TF-IDF, LIWC, ELMO, and BERT baselines.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"1907.02581","ref_index":25,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Transfer Learning for Risk Classification of Social Media Posts: Model Evaluation Study","primary_cat":"cs.CL","submitted_at":"2019-07-04T20:37:44+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":4.0,"formal_verification":"none","one_line_summary":"Finetuning GPT-1 on 150000 unlabeled Reachout.com posts then feeding the features into AutoML yields a new state-of-the-art macro F1 of 0.572 for triaging risk in 1588 labeled CLPsych 2017 posts without metadata or history.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"1906.08340","ref_index":4,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Learning Compressed Sentence Representations for On-Device Text Processing","primary_cat":"cs.CL","submitted_at":"2019-06-19T20:29:31+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"Four binarization strategies turn continuous sentence embeddings into binary form, cutting storage by over 98% with only about 2% performance drop on downstream tasks.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null}],"limit":50,"offset":0}