{"total":14,"items":[{"citing_arxiv_id":"2606.24320","ref_index":181,"ref_count":2,"confidence":0.9,"is_internal_anchor":false,"paper_title":"ZONOS2 Technical Report","primary_cat":"cs.SD","submitted_at":"2026-06-23T08:57:34+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":4.0,"formal_verification":"none","one_line_summary":"ZONOS2 8B is a scaled MoE TTS model with 900M active parameters trained on 6M hours of data that reports competitive SOTA results on naturalness, speaker similarity, WER, and a new ZTTS1-Eval benchmark while releasing weights and code.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.18089","ref_index":79,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"From Reasoning Traces to Reusable Modules: Understanding Compositional Generalization in Language Model Reasoning","primary_cat":"cs.LG","submitted_at":"2026-06-16T15:55:28+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Introduces a hierarchical latent selection model showing SFT supplies raw module materials in compound traces while RL decomposes them to identify atomic modules and enable recombination for new reasoning configurations.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.18307","ref_index":13,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"DRIFT: Refining Instruction Data via On-Policy Data Attribution","primary_cat":"cs.LG","submitted_at":"2026-06-16T07:21:49+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"DRIFT applies on-policy influence functions with signed weighting and debiasing to attribute and refine SFT data, raising performance on 7B instruction and reasoning models over prior curation methods.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.15893","ref_index":26,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"BALTO: Balanced Token-Level Policy Optimization for Hallucination Mitigation","primary_cat":"cs.CL","submitted_at":"2026-06-14T16:25:59+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"BALTO projects claim-level verification into balanced token-level rewards for RL-based hallucination mitigation in LLMs.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.07082","ref_index":10,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"On the Geometry of On-Policy Distillation","primary_cat":"cs.LG","submitted_at":"2026-06-05T09:20:15+00:00","verdict":"UNVERDICTED","verdict_confidence":"MODERATE","novelty_score":7.0,"formal_verification":"none","one_line_summary":"OPD updates occupy a relaxed off-principal regime and rapidly lock into a low-dimensional subspace that is functionally sufficient for its performance, distinct from SFT and RLVR trajectories.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.05559","ref_index":17,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"CLaaS: Continual learning as a service for sample efficient online learning","primary_cat":"cs.LG","submitted_at":"2026-06-04T01:14:19+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":4.0,"formal_verification":"none","one_line_summary":"CLaaS enables sample-efficient online continual learning for agents via replay-buffered parametric updates, outperforming in-context learning in forward transfer and retention on an adversarial task.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.01249","ref_index":236,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Trust Region On-Policy Distillation","primary_cat":"cs.LG","submitted_at":"2026-05-31T14:04:51+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"TrOPD stabilizes on-policy distillation for LLMs with trust-region learning, outlier estimation, and off-policy guidance, outperforming prior OPD methods on reasoning and code benchmarks.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.28819","ref_index":5,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"PEFT-Arena: Understanding Parameter-Efficient Finetuning from a Stability-Plasticity Perspective","primary_cat":"cs.LG","submitted_at":"2026-05-27T17:59:51+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"PEFT-Arena reveals distinct stability-plasticity profiles across PEFT methods, with orthogonal finetuning achieving the best Pareto frontier under comparable parameter budgets, supported by weight-space spectral and activation-space retention analyses.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.22005","ref_index":9,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Check Your LLM's Secret Dictionary! Five Lines of Code Reveal What Your LLM Learned (Including What It Shouldn't Have)","primary_cat":"cs.LG","submitted_at":"2026-05-21T05:02:51+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"SVD on the lm_head weight matrix of transformers reveals interpretable vocabulary clusters that indicate training data composition, model differences, and ethical concerns in models like GPT-OSS, Gemma, and Qwen.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.10973","ref_index":27,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Rotation-Preserving Supervised Fine-Tuning","primary_cat":"cs.LG","submitted_at":"2026-05-08T20:20:05+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"RPSFT improves the in-domain versus out-of-domain performance trade-off during LLM supervised fine-tuning by penalizing rotations in pretrained singular subspaces as a proxy for loss-sensitive directions.","context_count":1,"top_context_role":"background","top_context_polarity":"background","context_text":"4(Existence of an upper rank boundary).Assume there existsqsuch that cs = 0for alls /∈S q.(25) That is, all OOD-sensitive coordinates are already contained in the protected top-q block. Define the scalarized utility Φ(k) :=G id(k)−βF ood(k), β >0,(26) where β controls how strongly OOD forgetting is penalized relative to ID gain. Then every maximizer k⋆ ofΦsatisfies k⋆ ≤q.(27) Proof. For k≥q , enlarging the protected set no longer changes Food(k), because every coordinate withc s >0is already protected at rankq. Hence Food(k) =F ood(q), k≥q.(28) On the other hand, protecting any additional coordinate weakly decreases its contribution to Gid(k), with strict decrease wheneverλ >0andg s ̸= 0. Therefore Gid(k)≤G id(q), k≥q,(29)"},{"citing_arxiv_id":"2605.05365","ref_index":156,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"ZAYA1-8B Technical Report","primary_cat":"cs.AI","submitted_at":"2026-05-06T18:44:08+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"ZAYA1-8B is a reasoning MoE model with 700M active parameters that matches larger models on math and coding benchmarks and reaches 91.9% on AIME'25 via Markovian RSA test-time compute.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.20244","ref_index":33,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Hybrid Policy Distillation for LLMs","primary_cat":"cs.CL","submitted_at":"2026-04-22T06:46:22+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Hybrid Policy Distillation unifies existing knowledge distillation methods for LLMs into a reweighted log-likelihood objective and introduces a hybrid forward-reverse KL approach with mixed data sampling to improve stability, efficiency, and performance.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.17928","ref_index":54,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"HEALing Entropy Collapse: Enhancing Exploration in Few-Shot RLVR via Hybrid-Domain Entropy Dynamics Alignment","primary_cat":"cs.LG","submitted_at":"2026-04-20T08:09:01+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"HEAL mitigates entropy collapse in few-shot RLVR by selectively adding general-domain data and aligning trajectory-level entropy dynamics, matching full-shot performance with 32 target samples.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2509.25758","ref_index":22,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Thinking Sparks!: Emergent Attention Heads in Reasoning Models During Post Training","primary_cat":"cs.AI","submitted_at":"2025-09-30T04:23:43+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Post-training on reasoning tasks sparks the emergence of specialized attention heads that enable structured computation, with SFT adding stable heads while GRPO uses dynamic activation and pruning tied to reward signals, and controllable think models relying on compensatory heads instead of specific","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null}],"limit":50,"offset":0}