{"total":13,"items":[{"citing_arxiv_id":"2606.17628","ref_index":70,"ref_count":1,"confidence":0.88,"is_internal_anchor":false,"paper_title":"OPD-Evolver: Cultivating Holistic Agent Evolver via On-Policy Distillation","primary_cat":"cs.CL","submitted_at":"2026-06-16T07:33:53+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"OPD-Evolver uses on-policy self-distillation in fast interaction and slow attribution loops to build agents with holistic memory competence, outperforming prior systems by up to 11.5% and allowing a 9B model to compete with much larger ones.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.12838","ref_index":1,"ref_count":1,"confidence":0.88,"is_internal_anchor":false,"paper_title":"OCOO-T : A Simple and Scalable Virtual Cell Model for Transcriptional Perturbation Response Prediction","primary_cat":"q-bio.QM","submitted_at":"2026-06-11T03:04:38+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"OCOO-T is a flow-matching Transformer model that directly denoises continuous gene expression profiles to predict transcriptional responses to perturbations and reports state-of-the-art results on Tahoe100M, Replogle, and PBMC benchmarks.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.08816","ref_index":6,"ref_count":1,"confidence":0.88,"is_internal_anchor":false,"paper_title":"Knowledge Graphs and Reasoning LLMs for Finding Simple Yet Effective Transcriptomic Perturbation Predictors","primary_cat":"cs.LG","submitted_at":"2026-06-07T20:09:30+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"K-nearest neighbor from a knowledge graph beats most methods on out-of-distribution transcriptomic perturbation prediction, and an RL-trained reasoning LLM matches SOTA on Replogle et al. (2022) cell lines while improving downstream differential expression prediction.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.00685","ref_index":5,"ref_count":1,"confidence":0.88,"is_internal_anchor":false,"paper_title":"Prior-Guided Multi-Omic Transformers for Single-Cell Gene Regulatory Network Inference","primary_cat":"cs.LG","submitted_at":"2026-05-30T11:49:21+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"EpiAwareNet is a prior-guided multi-omic Transformer that uses gene-peak cross-attention for adaptive accessibility aggregation and bulk GRN priors for weak supervision to improve single-cell GRN reconstruction over baselines.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.24244","ref_index":25,"ref_count":1,"confidence":0.88,"is_internal_anchor":false,"paper_title":"MEDAL: Manifold Embedding Distillation via Autoencoder Learning","primary_cat":"stat.ML","submitted_at":"2026-05-22T21:45:12+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"MEDAL distills manifold embeddings into autoencoders to enable out-of-sample extension and held-out validation of dimension reduction methods.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.18576","ref_index":6,"ref_count":1,"confidence":0.88,"is_internal_anchor":false,"paper_title":"scHelix: Asymmetric Dual-Stream Integration via Explicit Gene-Level Disentanglement","primary_cat":"cs.LG","submitted_at":"2026-05-18T15:55:13+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"scHelix uses explicit gene-level partitioning into Anchors and Variants plus an asymmetric Align-Refine-Fuse dual-stream architecture to improve batch correction in scRNA-seq without over-correcting biological signals.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.06728","ref_index":8,"ref_count":1,"confidence":0.88,"is_internal_anchor":false,"paper_title":"OmicsLM: A Multimodal Large Language Model for Multi-Sample Omics Reasoning","primary_cat":"q-bio.GN","submitted_at":"2026-05-07T11:27:11+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"OmicsLM integrates continuous omics embeddings into LLMs for multi-sample biological reasoning, matching specialized models on profile tasks while outperforming them and general LLMs on language-guided QA over real expression data.","context_count":1,"top_context_role":"background","top_context_polarity":"unclear","context_text":"\", \"For directional claims in the expected output, verify that the \" \"actual output states the same direction.\", \"Evaluate whether the biological interpretation and mechanistic \" \"reasoning are consistent with the expected output.\", \"Assess completeness: missing the central finding or conclusion \" \"is a significant deficiency.\", ] DEEPEVAL_RUBRIC = [ Rubric(score_range=(0, 2), expected_outcome=\"Irrelevant, unanswered, or wrong subject.\"), Rubric(score_range=(3, 4), expected_outcome=\"Correct topic but major factual errors.\"), Rubric(score_range=(5, 6), expected_outcome=\"Partially correct but incomplete or inaccurate.\"), Rubric(score_range=(7, 8), expected_outcome=\"Main finding and most entities correct.\"), Rubric(score_range=(9, 10), expected_outcome=\"All key entities, directions, and conclusions align.\"), ] geval = GEval( 17 name=\"DeepEval\", evaluation_steps=DEEPEVAL_EVALUATION_STEPS, rubric=DEEPEVAL_RUBRIC, evaluation_params=[ LLMTestCaseParams.INPUT, LLMTestCaseParams.ACTUAL_OUTPUT, LLMTestCaseParams.EXPECTED_OUTPUT, ], model=GeminiModel(model=\"gemini-3.1-flash-lite-preview\"), ) A.3.3 Perturbation Prediction Perturbation prediction is a core capability of Cell2Sentence-Scale [Levine et al."},{"citing_arxiv_id":"2605.02142","ref_index":10,"ref_count":1,"confidence":0.88,"is_internal_anchor":false,"paper_title":"ORBIT: Learning Gene Program Co-Activation Structure for Cell-Type-Stratified Pathway Rewiring Analysis in Single-Cell Transcriptomics","primary_cat":"q-bio.GN","submitted_at":"2026-05-04T01:50:51+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"ORBIT uses an intervention-consistent self-supervised objective in a transformer to infer asymmetric gene program influences from observational scRNA-seq data, recovering Alzheimer's vulnerability patterns and achieving 0.984 macro F1 cell-type classification from 220 pathway scores.","context_count":1,"top_context_role":"baseline","top_context_polarity":"baseline","context_text":"Statistical significance is assessed by permuting condition labels 1,000 times with Benjamini-HochbergFDRcorrectionperdirectedpair. Allreportedpairswith |∆ ¯A| ≥0.004 survive FDR atq <0.05across all three vocabularies. Gene-level projection.To connect program-level rewiring to individual genes, we define a gene- pathway influence matrix: Ig,p2 = X p1 Mg,p1 · ¯Ap1,p2 ,(10) where M∈ {0,1} G×P is the binary gene-program membership matrix. Substituting∆ ¯A yields ∆I, identifying genes most implicated in rewiring. 3 Experiments 3.1 Data and Implementation Data.Stage 1 uses the Morabito 2021 snRNA-seq atlas (GSE174367; 191,890 nuclei, seven cell types, prefrontal cortex) [Morabito et al., 2021]. Cross-dataset replication uses a held-out disease"},{"citing_arxiv_id":"2605.08128","ref_index":11,"ref_count":1,"confidence":0.88,"is_internal_anchor":false,"paper_title":"Towards Universal Gene Regulatory Network Inference: Unlocking Generalizable Regulatory Knowledge in Single-cell Foundation Models","primary_cat":"cs.LG","submitted_at":"2026-05-01T01:45:24+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"Two new methods distill implicit regulatory knowledge from single-cell foundation models to enable generalizable gene regulatory network inference on unseen data.","context_count":1,"top_context_role":"background","top_context_polarity":"background","context_text":"Methods, 22(10):1657-1661, 2025. doi: 10.1038/ s41592-025-02772-6. URL https://doi.org/10. 1038/s41592-025-02772-6. Aibar, S., Gonz'alez-Blas, C. B., Moerman, T., Huynh-Thu, V . A., Imrichov'a, H., Hulselmans, G., Rambow, J.-C., Marine, J.-C., Geurts, P., Aerts, J., et al. Scenic: single- cell regulatory network inference and clustering.Nature Methods, 14(11):1083-1086, 2017. doi: 10.1038/nmeth. 4463. Barab'asi, A.-L. and Oltvai, Z. N. Network biology: understanding the cell's functional organization.Na- ture Reviews Genetics, 5(2):101-113, 2004. doi: 10. 1038/nrg1272. URL https://doi.org/10.1038/ nrg1272. Browaeys, R., Saelens, W., and Saeys, Y . Nichenet: modeling intercellular communication by linking lig-"},{"citing_arxiv_id":"2604.16642","ref_index":7,"ref_count":2,"confidence":0.88,"is_internal_anchor":false,"paper_title":"Geometric coherence of single-cell CRISPR perturbations reveals regulatory architecture and predicts cellular stress","primary_cat":"q-bio.QM","submitted_at":"2026-04-17T19:01:05+00:00","verdict":null,"verdict_confidence":null,"novelty_score":null,"formal_verification":null,"one_line_summary":null,"context_count":1,"top_context_role":"background","top_context_polarity":"unclear","context_text":"doi:10.1074/mcp. m700510-mcp200.url:http://dx.doi.org/10.1074/mcp.M700510-MCP200. [6] John D. Crispino and Mitchell J. Weiss. \"Erythro-megakaryocytic transcription factors associated with hereditary anemia\". In:Blood123.20 (May 2014), pp. 3080-3088.issn: 1528-0020.doi:10.1182/blood-2014-01-453167.url: http://dx.doi.org/10.1182/blood-2014-01-453167. [7] Haotian Cui, Chloe Wang, Hassaan Maan, Kuan Pang, Fengning Luo, Nan Duan, and Bo Wang. \"scGPT: toward building a foundation model for single-cell multi-omics using generative AI\". In:Nature Methods21.8 (Feb. 2024), pp. 1470-1480.issn: 1548-7105.doi:10.1038/s41592-024-02201-0. [8] Jose Davila-Velderrain, Juan C. Martinez-Garcia, and Elena R. Alvarez-Buylla."},{"citing_arxiv_id":"2603.00678","ref_index":11,"ref_count":1,"confidence":0.88,"is_internal_anchor":false,"paper_title":"From Syntax to Semantics: Geometric Stability as the Missing Axis of Perturbation Biology","primary_cat":"q-bio.QM","submitted_at":"2026-02-28T14:42:50+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Geometric stability, defined as the directional coherence of cellular responses to perturbation, provides a framework for assessing whether resulting cellular states are stable beyond conventional metrics of intervention success.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2602.00586","ref_index":28,"ref_count":1,"confidence":0.88,"is_internal_anchor":false,"paper_title":"RAG-GNN: Integrating Retrieved Knowledge with Graph Neural Networks for Precision Medicine","primary_cat":"q-bio.MN","submitted_at":"2026-01-31T08:05:02+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"RAG-GNN augments GNNs with retrieved literature knowledge via gated fusion to improve functional clustering of 379 proteins in cancer signaling networks, raising silhouette score by 0.093.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2505.11771","ref_index":7,"ref_count":1,"confidence":0.88,"is_internal_anchor":false,"paper_title":"Residual Feature Integration is Sufficient to Prevent Negative Transfer","primary_cat":"cs.LG","submitted_at":"2025-05-17T00:36:59+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"Residual feature integration with a trainable target-side encoder provably prevents negative transfer, achieving convergence rates no worse than training from scratch under informative target distributions.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null}],"limit":50,"offset":0}