# paper-forge daily report — 2026-10-14

178 new papers screened; 10 analyzed in depth; 0 reproduction runs.

## Reproductions

- none today (reimplementation is opt-in: `forge daily --reimplement`)

## Top candidates

### [Gaussian Process Bandits for AI System Optimization with Selectively Verified Feedback](https://arxiv.org/abs/2610.17760)
`2610.17760` · score 63.7 · analysis: heuristic · feasibility 8/10

Gaussian process bandits exploit similarities among configurations to optimize prompts, retrieval systems, and agent workflows. In these applications, however, inexpensive model-based evaluations can disagree with trusted task outcomes.

Rubric evidence: + toy/synthetic experiments ("Synthetic"); + small/synthetic data ("Synthetic"); + explicit algorithm/pseudocode ("Algorithm 1"); + formal statement ("bound"); no public code found; + keywords: retrieval, kernel, bandit

### [Online LLM Personalization from Pairwise Preferences Across Users](https://arxiv.org/abs/2610.17203)
`2610.17203` · score 63.7 · analysis: heuristic · feasibility 7/10

Large language models can answer the same question in different styles, but users' preferred styles are hidden and vary across users and questions. We study inference-time personalization through online selection from a fixed set of prompts, each representing a persona.

Rubric evidence: - LLM-based ("LLM"); + toy/synthetic experiments ("simulated"); + small/synthetic data ("simulated"); + public benchmarks ("benchmark"); + concrete method ("closed form"); + formal statement ("bound"); + claims simplicity ("simple"); + code: https://github.com/RoGy-Labs/tutorduel-code; + keywords: bandit

### [SCOPE Before You Sample: Conformal Scores Guide Posterior Exploration for Fast and Calibrated Bayesian Predictive Inference](https://arxiv.org/abs/2610.17146)
`2610.17146` · score 61.6 · analysis: heuristic · feasibility 8/10

Bayesian predictive regions account for parameter uncertainty but can be poorly calibrated under model misspecification, even with expensive full MCMC. Cheaper approximations such as variational inference reduce the cost but often yield less accurate predictive distributions.

Rubric evidence: + toy/synthetic experiments ("simulations"); + small/synthetic data ("simulated"); + public benchmarks ("benchmarks"); + explicit algorithm/pseudocode ("Algorithm 1"); + formal statement ("bound"); no public code found; + keywords: monte carlo, variational

### [Exact Topological Compliance: Generating Persistence-Equivalent Graphs](https://arxiv.org/abs/2610.18856)
`2610.18856` · score 60.0 · analysis: heuristic · feasibility 6/10

Persistent homology summarizes the topological evolution of a graph as persistence diagrams (PDs). Generating graphs whose topology matches a target descriptor is increasingly important, yet existing topology-aware generative models rely on soft regularization, producing graphs whose PDs may only be approximately similar.

Rubric evidence: + public benchmarks ("benchmarks"); + explicit algorithm/pseudocode ("Algorithm 1"); + formal statement ("we prove"); + claims simplicity ("simple"); + code: https://github.com/Aalto-QuML/ExactCompliance; + keywords: regularization

### [Minimax Optimality of Shampoo for Nonstationary Stochastic Optimization](https://arxiv.org/abs/2610.16923)
`2610.16923` · score 57.4 · analysis: heuristic · feasibility 8/10

We develop a comprehensive theoretical analysis of Matrix Shampoo for tracking time-varying optima in nonstationary stochastic optimization. We derive finite-horizon high-probability bounds for average tracking error that disentangle initialization, optimizer drift, momentum-induced drift penalty, stochastic noise, and adaptive preconditioning, with explicit dependence on momentum parameter \(β_1\) and stabilization parameter \(\varepsilon\).

Rubric evidence: + toy/synthetic experiments ("synthetic"); + small/synthetic data ("synthetic"); + explicit algorithm/pseudocode ("Algorithm 1"); + formal statement ("bound"); no public code found; + keywords: optimizer

### [JevGPT: Thinking Fast and Slow in Autonomous Research](https://arxiv.org/abs/2610.18272)
`2610.18272` · score 57.1 · analysis: heuristic · feasibility 5/10

Autonomous research agents must acquire and reason over evidence within a limited budget. Yet relevant sources can remain uninspected or unused, leaving a gap between retrieval coverage and the evidence that informs research.

Rubric evidence: + formal statement ("bound"); + code: https://github.com/sunyasheng/JevGPT; + keywords: retrieval, ranking; + 6 HF upvotes

### [PsyBer-Agent: A Psychology-Driven User Agent for Recommender Systems](https://arxiv.org/abs/2610.13989)
`2610.13989` · score 56.3 · analysis: heuristic · feasibility 6/10

Evaluating recommender systems before deployment requires simulators that remain realistic over long, closed-loop interaction cycles. Current LLM-based agents generally function as surface imitators that generate realistic short-term responses but fail to capture the underlying user dynamics such as fatigue, affect, and shifting preferences.

Rubric evidence: - LLM-based ("LLM"); + toy/synthetic experiments ("simulated"); + CPU-friendly ("lightweight"); + small/synthetic data ("simulated"); + public benchmarks ("Benchmark"); - system/framework (many moving parts) ("framework"); + code: https://github.com/franz-chang/PsyBer-Agent; + 4 HF upvotes

### [Joint Neural Posterior Estimation and Model Selection for Chemical Reaction Networks](https://arxiv.org/abs/2610.18825)
`2610.18825` · score 54.2 · analysis: heuristic · feasibility 5/10

Simulation-based Bayesian inference for Chemical Reaction Networks (CRNs) must simultaneously solve two computationally demanding problems: discriminating between overlapping reaction pathways and estimating reaction rate constants. Classical approaches, such as Sequential Monte Carlo ABC, rely on handcrafted summary statistics and scale poorly across large parameter spaces.

Rubric evidence: + toy/synthetic experiments ("Simulation"); + formal statement ("theorem"); + code: https://github.com/Esquivel-Arturo/JNPE; + keywords: monte carlo

### [FASHION-RQ: Mining Hierarchical Semantic Structure for Automated Evaluation of Fashion Retrieval](https://arxiv.org/abs/2610.17841)
`2610.17841` · score 53.2 · analysis: heuristic · feasibility 6/10

Evaluating fashion retrieval systems remains challenging due to the limited availability of benchmarks that capture the fine-grained, open-ended queries encountered in real-world search and recommendation. Existing benchmarks often rely on manual curation, predefined taxonomies, or synthetic prompts, limiting scalability and semantic coverage.

Rubric evidence: + toy/synthetic experiments ("synthetic"); + small/synthetic data ("synthetic"); + public benchmarks ("benchmarks"); - system/framework (many moving parts) ("framework"); no public code found; + keywords: retrieval, variational, ranking

### [SWE-MISBEHAVIOR: Benchmarking and Mitigating Boundary Violations in Software Coding Agents](https://arxiv.org/abs/2610.18206)
`2610.18206` · score 49.9 · analysis: heuristic · feasibility 4/10

We present SWE-MISBEHAVIOR, the first taxonomy and evaluation framework for diagnosing boundary-violating misbehaviors in coding agents. It contains 7.4K human-in-the-loop synthesized samples spanning three layers of agent action space: (i) Code Generation, (ii) File Management, and (iii) System Interaction.

Rubric evidence: + CPU-friendly ("lightweight"); - system/framework (many moving parts) ("framework"); + code: https://github.com/dxlong2000/SWE-Misbehavior; + 5 HF upvotes

## All screened papers

| score | paper | compute | data | clarity | code | interest | analysis | laptop-feasible |
|---:|---|---:|---:|---:|---:|---:|---|---|
| 63.7 | [Gaussian Process Bandits for AI System Optimization with Selectively Verified Feedback](https://arxiv.org/abs/2610.17760) `2610.17760` | 0.65 | 0.75 | 0.85 | 0.00 | 0.60 | heuristic | True |
| 63.7 | [Online LLM Personalization from Pairwise Preferences Across Users](https://arxiv.org/abs/2610.17203) `2610.17203` | 0.45 | 0.80 | 0.90 | 1.00 | 0.20 | heuristic | False |
| 61.6 | [SCOPE Before You Sample: Conformal Scores Guide Posterior Exploration for Fast and Calibrated Bayesian Predictive Inference](https://arxiv.org/abs/2610.17146) `2610.17146` | 0.65 | 0.80 | 0.85 | 0.00 | 0.40 | heuristic | True |
| 60.0 | [Exact Topological Compliance: Generating Persistence-Equivalent Graphs](https://arxiv.org/abs/2610.18856) `2610.18856` | 0.50 | 0.45 | 1.00 | 1.00 | 0.20 | heuristic | True |
| 57.4 | [Minimax Optimality of Shampoo for Nonstationary Stochastic Optimization](https://arxiv.org/abs/2610.16923) `2610.16923` | 0.65 | 0.75 | 0.85 | 0.00 | 0.20 | heuristic | True |
| 57.1 | [JevGPT: Thinking Fast and Slow in Autonomous Research](https://arxiv.org/abs/2610.18272) `2610.18272` | 0.50 | 0.40 | 0.55 | 1.00 | 0.68 | heuristic | True |
| 56.3 | [PsyBer-Agent: A Psychology-Driven User Agent for Recommender Systems](https://arxiv.org/abs/2610.13989) `2610.13989` | 0.60 | 0.80 | 0.30 | 1.00 | 0.23 | heuristic | True |
| 54.2 | [Joint Neural Posterior Estimation and Model Selection for Chemical Reaction Networks](https://arxiv.org/abs/2610.18825) `2610.18825` | 0.65 | 0.40 | 0.55 | 1.00 | 0.20 | heuristic | True |
| 53.2 | [FASHION-RQ: Mining Hierarchical Semantic Structure for Automated Evaluation of Fashion Retrieval](https://arxiv.org/abs/2610.17841) `2610.17841` | 0.65 | 0.80 | 0.30 | 0.00 | 0.60 | heuristic | True |
| 49.9 | [SWE-MISBEHAVIOR: Benchmarking and Mitigating Boundary Violations in Software Coding Agents](https://arxiv.org/abs/2610.18206) `2610.18206` | 0.65 | 0.40 | 0.30 | 1.00 | 0.26 | heuristic | True |
| 48.9 | [Graph-free Retrieval with Graph-Modified Embeddings](https://arxiv.org/abs/2610.18410) `2610.18410` | 0.65 | 0.80 | 0.40 | 0.00 | 0.20 | — | — |
| 48.9 | [CLASSO: a practical primal path-following algorithm for the constrained LASSO](https://arxiv.org/abs/2610.17024) `2610.17024` | 0.65 | 0.80 | 0.55 | 0.00 | 0.00 | — | — |
| 48.9 | [Sample-Size Extrapolation in Simulation-Based Inference via Statistical Scaling](https://arxiv.org/abs/2610.16820) `2610.16820` | 0.85 | 0.45 | 0.60 | 0.00 | 0.00 | — | — |
| 47.9 | [Weighting-Based Strategies for Classification: Analysis and Local Peak Dynamics with CART Trees](https://arxiv.org/abs/2610.18037) `2610.18037` | 0.65 | 0.75 | 0.55 | 0.00 | 0.00 | — | — |
| 47.9 | [The Surprising Expressivity of a Single Neuron with Autoregressive Chain-of-Thought](https://arxiv.org/abs/2610.17283) `2610.17283` | 0.65 | 0.75 | 0.55 | 0.00 | 0.00 | — | — |
| 47.9 | [Designing for Prediction: Planning Formulas for Expected Out-of-Sample $R^2$ and $r$](https://arxiv.org/abs/2610.18781) `2610.18781` | 0.85 | 0.40 | 0.60 | 0.00 | 0.00 | — | — |
| 47.9 | [ReTerm: Recovery Training for Self-Improving Terminal Agents](https://arxiv.org/abs/2610.18484) `2610.18484` | 0.50 | 0.40 | 0.40 | 1.00 | 0.30 | — | — |
| 47.4 | [Two Clocks: Why Larger Rollout Batches Learn Slower but Fail on Schedule in Reinforcement Learning with Verifiable Rewards](https://arxiv.org/abs/2610.18853) `2610.18853` | 0.50 | 0.40 | 0.45 | 1.00 | 0.20 | — | — |
| 47.4 | [KNN Implementation Details Can Dramatically Change Performance: An Example from Cover Trees](https://arxiv.org/abs/2610.16940) `2610.16940` | 0.50 | 0.45 | 0.40 | 1.00 | 0.20 | — | — |
| 47.4 | [Risk-Averse Inference-Time Diffusion Model Alignment with Finite-Sample Guarantees](https://arxiv.org/abs/2610.17579) `2610.17579` | 0.50 | 0.75 | 0.45 | 0.00 | 0.40 | — | — |
| 46.8 | [Convergence-Rate-Guided Algorithm Design of Expected Improvement for Continuous Simulation Optimization](https://arxiv.org/abs/2610.18434) `2610.18434` | 0.65 | 0.40 | 0.70 | 0.00 | 0.20 | — | — |
| 46.3 | [Enabling Token-wise Importance Sampling in Off-Policy RLVR with Proximal KL Regularization](https://arxiv.org/abs/2610.18738) `2610.18738` | 0.50 | 0.45 | 0.55 | 0.00 | 0.60 | — | — |
| 46.3 | [Follow the Perturbed Leader for Linear Bandits: $\sqrt{T}$ Regret over General Convex Sets](https://arxiv.org/abs/2610.17818) `2610.17818` | 0.70 | 0.40 | 0.60 | 0.00 | 0.20 | — | — |
| 45.8 | [Hierarchical Time Series Forecasting with Operator-Valued Kernels](https://arxiv.org/abs/2610.18196) `2610.18196` | 0.65 | 0.75 | 0.30 | 0.00 | 0.20 | — | — |
| 45.3 | [EHR-Bench: A Chinese Multi-Task Benchmark for Recruitment AI](https://arxiv.org/abs/2610.16689) `2610.16689` | 0.30 | 0.45 | 0.30 | 1.00 | 0.60 | — | — |
| 44.7 | [WeGR: Unified Generative Retrieval for Industrial Search Systems in WeChat](https://arxiv.org/abs/2610.16979) `2610.16979` | 0.65 | 0.40 | 0.30 | 0.00 | 0.60 | — | — |
| 44.7 | [Monitoring for Shifts at Unknown Times: Parameter-Free Conditional Conformal Martingales](https://arxiv.org/abs/2610.17343) `2610.17343` | 0.65 | 0.75 | 0.40 | 0.00 | 0.00 | — | — |
| 44.7 | [Tight End-to-End Accounting for Differentially Private Synthetic Data](https://arxiv.org/abs/2610.17249) `2610.17249` | 0.65 | 0.75 | 0.40 | 0.00 | 0.00 | — | — |
| 44.7 | [Self-Distillation with Verbal Regularization](https://arxiv.org/abs/2610.16354) `2610.16354` | 0.30 | 0.40 | 0.40 | 1.00 | 0.50 | — | — |
| 44.2 | [Self-Evolving Agent Harness](https://arxiv.org/abs/2610.18813) `2610.18813` | 0.50 | 0.40 | 0.45 | 1.00 | 0.00 | — | — |
| 43.7 | [SEAR: Self-Evolving Agentic Robotics](https://arxiv.org/abs/2610.18792) `2610.18792` | 0.65 | 0.80 | 0.30 | 0.00 | 0.00 | — | — |
| 43.7 | [ProbabilityBench: Model Evaluation and Calibration with Exact Ground-Truth Uncertainty](https://arxiv.org/abs/2610.17016) `2610.17016` | 0.65 | 0.80 | 0.30 | 0.00 | 0.00 | — | — |
| 43.2 | [Radial Tail Repair: Post-Training Correction for Diffusion and Flow Models](https://arxiv.org/abs/2610.16851) `2610.16851` | 0.40 | 0.75 | 0.55 | 0.00 | 0.20 | — | — |
| 43.2 | [Learning Source Structure: Bayesian Prediction with Transformers](https://arxiv.org/abs/2610.18725) `2610.18725` | 0.50 | 0.40 | 0.40 | 1.00 | 0.00 | — | — |
| 43.2 | [Deterministic Prediction as Global Velocity: Flow Matching for Probabilistic Time-Series Forecast](https://arxiv.org/abs/2610.18621) `2610.18621` | 0.50 | 0.40 | 0.40 | 1.00 | 0.00 | — | — |
| 42.3 | [HexaBee-4B: Six-Language Translation with Multilingual Answering and Bidirectional Supervision](https://arxiv.org/abs/2610.16361) `2610.16361` | 0.00 | 0.75 | 0.40 | 1.00 | 0.48 | — | — |
| 42.1 | [Cepheid: Decoupling Scale from Structure for Cross-Family Motion Completion](https://arxiv.org/abs/2610.18871) `2610.18871` | 0.65 | 0.40 | 0.40 | 0.00 | 0.30 | — | — |
| 42.1 | [Fast Rates for Offline Q-Learning with KL Regularization](https://arxiv.org/abs/2610.18839) `2610.18839` | 0.50 | 0.40 | 0.70 | 0.00 | 0.20 | — | — |
| 42.1 | [Knowledge Graphs as Files: An Agent Harness Where Tool Creation and Tool Use Co-Evolve](https://arxiv.org/abs/2610.18616) `2610.18616` | 0.50 | 0.45 | 0.30 | 1.00 | 0.00 | — | — |
| 41.6 | [Knowing the User Is Not Knowing What We Established: Grounding Awareness in Personal Agents](https://arxiv.org/abs/2610.18684) `2610.18684` | 0.65 | 0.45 | 0.40 | 0.00 | 0.20 | — | — |
| 41.1 | [Who Improves the Robot? Toward Physical Recursive Self-Improvement](https://arxiv.org/abs/2610.18842) `2610.18842` | 0.50 | 0.40 | 0.30 | 1.00 | 0.00 | — | — |
| 40.5 | [Hybrid Reasoning Agents](https://arxiv.org/abs/2610.18861) `2610.18861` | 0.45 | 0.45 | 0.30 | 1.00 | 0.00 | — | — |
| 40.5 | [GeoReader: Learning Latent Geometry for Spatial Reasoning in Vision-Language Models](https://arxiv.org/abs/2610.18694) `2610.18694` | 0.45 | 0.45 | 0.30 | 1.00 | 0.00 | — | — |
| 40.5 | [MultiAlloc: Budget-Aware Precision Allocation for KV Cache Quantization](https://arxiv.org/abs/2610.18642) `2610.18642` | 0.45 | 0.45 | 0.30 | 1.00 | 0.00 | — | — |
| 40.5 | [Agent-Reflex: Teaching Language Agents to Act Reflectively](https://arxiv.org/abs/2610.17953) `2610.17953` | 0.30 | 0.45 | 0.30 | 1.00 | 0.30 | — | — |
| 40.2 | [Elephant in the Fridge: Constant-Memory Frame Packing for Long Video Understanding](https://arxiv.org/abs/2610.16118) `2610.16118` | 0.20 | 0.45 | 0.45 | 1.00 | 0.28 | — | — |
| 40.0 | [Measure Neural Mappings](https://arxiv.org/abs/2610.17384) `2610.17384` | 0.50 | 0.45 | 0.55 | 0.00 | 0.20 | — | — |
| 39.5 | [Structured SDEs for Realized GARCH: Continuous-Time Volatility Modelling and Component-Level Analysis](https://arxiv.org/abs/2610.17732) `2610.17732` | 0.65 | 0.45 | 0.30 | 0.00 | 0.20 | — | — |
| 39.5 | [SHELDON: Reasoning through Dual Search over Hypothesis and Observation Spaces](https://arxiv.org/abs/2610.18025) `2610.18025` | 0.30 | 0.45 | 0.40 | 1.00 | 0.10 | — | — |
| 38.9 | [A Refresh Interval Does Not Fully Specify a Refresh Schedule: Learning-Rate-Dependent Placement in Kronecker-Preconditioned Language Model Training](https://arxiv.org/abs/2610.18695) `2610.18695` | 0.50 | 0.40 | 0.40 | 0.00 | 0.40 | — | — |
| 38.9 | [Retrieval Disagreement as a Signal: Diagnosing and Steering Multilingual Dense Embedders](https://arxiv.org/abs/2610.18526) `2610.18526` | 0.50 | 0.40 | 0.40 | 0.00 | 0.40 | — | — |
| 38.9 | [Learning a Group Element from Its Action: Galois Correspondences, Optimal ERM, and an Abel-Ruffini Threshold for the Natarajan Dimension](https://arxiv.org/abs/2610.18154) `2610.18154` | 0.50 | 0.40 | 0.70 | 0.00 | 0.00 | — | — |
| 38.9 | [Reliable Scale-Free Bandits via Empirical Likelihood](https://arxiv.org/abs/2610.18055) `2610.18055` | 0.50 | 0.40 | 0.55 | 0.00 | 0.20 | — | — |
| 38.9 | [Accounting for model error in ensemble variational data assimilation through learned generative dynamics](https://arxiv.org/abs/2610.17333) `2610.17333` | 0.50 | 0.40 | 0.40 | 0.00 | 0.40 | — | — |
| 38.4 | [How Safe Is ChatGPT Health Advice? Evidence from Real User Conversations in Four Global South Countries](https://arxiv.org/abs/2610.18465) `2610.18465` | 0.45 | 0.75 | 0.40 | 0.00 | 0.00 | — | — |
| 38.4 | [SimInter: Evidence-Grounded User Simulation with Interview-Based Profile Enrichment for Conversational Recommender Systems](https://arxiv.org/abs/2610.18308) `2610.18308` | 0.45 | 0.75 | 0.40 | 0.00 | 0.00 | — | — |
| 37.9 | [SkillScaffold: Scaffolding Agentic Reinforcement Learning with Policy-Adaptive Skill Credit Assignment and Selective Injection](https://arxiv.org/abs/2610.18807) `2610.18807` | 0.30 | 0.40 | 0.30 | 1.00 | 0.20 | — | — |
| 37.9 | [ProAS: A Progress-Aware Advantage Shaping Framework for Multi-Turn Agentic Reinforcement Learning](https://arxiv.org/abs/2610.18683) `2610.18683` | 0.30 | 0.40 | 0.45 | 1.00 | 0.00 | — | — |
| 37.9 | [Jev-ToM: Evaluating Theory of Mind in Typed Decision Models](https://arxiv.org/abs/2610.18610) `2610.18610` | 0.30 | 0.45 | 0.40 | 1.00 | 0.00 | — | — |
| 37.4 | [ndscape: a structural map of the nested dichotomy space](https://arxiv.org/abs/2610.18050) `2610.18050` | 0.65 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 37.1 | [Rubrics from the Task, Verdicts from Evidence: Pointwise Rubric Rewards with Tool-Augmented Judges](https://arxiv.org/abs/2610.18766) `2610.18766` | 0.50 | 0.40 | 0.40 | 0.00 | 0.28 | — | — |
| 36.8 | [BridgeTwist: Twisting Schrödinger Bridges for Training-Free Conditional Sampling](https://arxiv.org/abs/2610.18658) `2610.18658` | 0.30 | 0.40 | 0.60 | 0.00 | 0.40 | — | — |
| 36.8 | [AMemOps: Evaluating Memory Updates in Long Audio Streams](https://arxiv.org/abs/2610.18652) `2610.18652` | 0.50 | 0.45 | 0.40 | 0.00 | 0.20 | — | — |
| 36.8 | [Mitigating Popularity Bias in Sequential Recommendations with Sparse Prior-Regularized Objective](https://arxiv.org/abs/2610.17823) `2610.17823` | 0.50 | 0.45 | 0.40 | 0.00 | 0.20 | — | — |
| 36.8 | [Reference Wording Shapes Retrieval of Robot Task Instructions](https://arxiv.org/abs/2610.17549) `2610.17549` | 0.50 | 0.45 | 0.40 | 0.00 | 0.20 | — | — |
| 36.8 | [Computed, not Memorized: Bit-Factorized Identifiers for Efficient Generative Retrieval](https://arxiv.org/abs/2610.17467) `2610.17467` | 0.50 | 0.45 | 0.40 | 0.00 | 0.20 | — | — |
| 36.8 | [Counterfactual Certification for Sequential Unlearning with Long-Run Performance Guarantees](https://arxiv.org/abs/2610.17993) `2610.17993` | 0.70 | 0.40 | 0.30 | 0.00 | 0.00 | — | — |
| 36.8 | [Near-Optimal Regret Bounds for Monotone Stochastic Optimization under Semi-Bandit Feedback](https://arxiv.org/abs/2610.17320) `2610.17320` | 0.50 | 0.40 | 0.45 | 0.00 | 0.20 | — | — |
| 36.8 | [Measuring Variable Importance Using Proportional Inclusive Marginal Effects](https://arxiv.org/abs/2610.17053) `2610.17053` | 0.50 | 0.40 | 0.60 | 0.00 | 0.00 | — | — |
| 36.8 | [APEX: Asynchronous Prefix Denoising for Extreme Compression](https://arxiv.org/abs/2610.12843) `2610.12843` | 0.35 | 0.00 | 0.40 | 1.00 | 0.43 | — | — |
| 36.3 | [Choosing LLMs for Strategic Agent Interactions](https://arxiv.org/abs/2610.18354) `2610.18354` | 0.45 | 0.45 | 0.45 | 0.00 | 0.20 | — | — |
| 35.8 | [Statistical Vignettes on the Grokking Phenomenon](https://arxiv.org/abs/2610.18811) `2610.18811` | 0.50 | 0.40 | 0.55 | 0.00 | 0.00 | — | — |
| 35.8 | [A Rate Transfer Method: Convergence Rates in Stochastic Gradient Descent via Stochastic Calculus](https://arxiv.org/abs/2610.18739) `2610.18739` | 0.50 | 0.40 | 0.55 | 0.00 | 0.00 | — | — |
| 35.8 | [On Collusion-Free Teaching in Planar Graphs](https://arxiv.org/abs/2610.18714) `2610.18714` | 0.50 | 0.40 | 0.55 | 0.00 | 0.00 | — | — |
| 35.8 | [Minimax rates for pointwise CATE estimation: the role of design smoothness](https://arxiv.org/abs/2610.18623) `2610.18623` | 0.50 | 0.40 | 0.55 | 0.00 | 0.00 | — | — |
| 35.8 | [Pack Light, CarryOn: Building Object-Oriented Context for Long-Horizon Agents](https://arxiv.org/abs/2610.18784) `2610.18784` | 0.50 | 0.40 | 0.55 | 0.00 | 0.00 | — | — |
| 35.8 | [Known-Voice Cloning and the Limits of Personality Transfer for Character Design](https://arxiv.org/abs/2610.18669) `2610.18669` | 0.50 | 0.40 | 0.55 | 0.00 | 0.00 | — | — |
| 35.8 | [Estimating Surprisal from the Speech Signal](https://arxiv.org/abs/2610.18439) `2610.18439` | 0.50 | 0.40 | 0.40 | 0.00 | 0.20 | — | — |
| 35.8 | [Structural Segmentation for Brazilian Legal Document Retrieval](https://arxiv.org/abs/2610.18622) `2610.18622` | 0.50 | 0.40 | 0.40 | 0.00 | 0.20 | — | — |
| 35.8 | [AiSAQ-in-Filtering for DRAM-free Filtered Information Retrieval](https://arxiv.org/abs/2610.17627) `2610.17627` | 0.50 | 0.40 | 0.40 | 0.00 | 0.20 | — | — |
| 35.8 | [From Conceptual EneMonIE to an Executable Backend: A Modular Architecture for Energy Information Extractionand Grounded Question Answering](https://arxiv.org/abs/2610.17243) `2610.17243` | 0.50 | 0.40 | 0.40 | 0.00 | 0.20 | — | — |
| 35.8 | [From Fourier Amplitudes to Persistence Images for Impact-Echo Defect Detection](https://arxiv.org/abs/2610.18732) `2610.18732` | 0.50 | 0.40 | 0.40 | 0.00 | 0.20 | — | — |
| 35.8 | [Recalibrating Many Questions with Few Labels: Minimax Rates and Identification Limits](https://arxiv.org/abs/2610.18488) `2610.18488` | 0.50 | 0.40 | 0.55 | 0.00 | 0.00 | — | — |
| 35.8 | [On the Optimality of Random Features under General Conditions: Misspecified Cases and Robust Learning](https://arxiv.org/abs/2610.18333) `2610.18333` | 0.50 | 0.40 | 0.40 | 0.00 | 0.20 | — | — |
| 35.8 | [Lifting Finite-Sample Gaussian Approximation of Wide Neural Networks to Functional Convergence](https://arxiv.org/abs/2610.18166) `2610.18166` | 0.50 | 0.40 | 0.55 | 0.00 | 0.00 | — | — |
| 35.8 | [Learning Distributions from Samples and Qualities](https://arxiv.org/abs/2610.17956) `2610.17956` | 0.50 | 0.40 | 0.55 | 0.00 | 0.00 | — | — |
| 35.8 | [Effect of Spherical Parameter Space on Number of Non-Forgetting Tasks and Capacity of Continual Learning](https://arxiv.org/abs/2610.17453) `2610.17453` | 0.50 | 0.40 | 0.55 | 0.00 | 0.00 | — | — |
| 35.8 | [In-context classification: Calibration and noise in linear attention with softmax readout](https://arxiv.org/abs/2610.17061) `2610.17061` | 0.40 | 0.40 | 0.55 | 0.00 | 0.20 | — | — |
| 35.3 | [RoboSimStudio: An Interactive Toolkit for Rapid Real2Sim of Robot Manipulation Tasks](https://arxiv.org/abs/2610.18835) `2610.18835` | 0.65 | 0.40 | 0.30 | 0.00 | 0.00 | — | — |
| 35.3 | [Texture Interpolation as Texton Matching](https://arxiv.org/abs/2610.18655) `2610.18655` | 0.25 | 0.40 | 0.40 | 1.00 | 0.00 | — | — |
| 35.3 | [MixLID: Token-Level Language Identification in Mixed-Language Text](https://arxiv.org/abs/2610.16601) `2610.16601` | 0.50 | 0.45 | 0.40 | 0.00 | 0.10 | — | — |
| 34.7 | [Read the Room Before Your Model Generates](https://arxiv.org/abs/2610.18820) `2610.18820` | 0.30 | 0.40 | 0.65 | 0.00 | 0.20 | — | — |
| 34.7 | [Legal-R1: Agentic Retrieval for Evidence-Based Legal Reasoning](https://arxiv.org/abs/2610.18611) `2610.18611` | 0.50 | 0.45 | 0.30 | 0.00 | 0.20 | — | — |
| 34.7 | [HASTE: Hierarchical and Attribute-aware Supervision for Generative Recommendation](https://arxiv.org/abs/2610.17368) `2610.17368` | 0.50 | 0.45 | 0.45 | 0.00 | 0.00 | — | — |
| 33.7 | [On the Monitorability of Natural Language Autoencoders: Identification vs Discrimination](https://arxiv.org/abs/2610.18727) `2610.18727` | 0.50 | 0.40 | 0.45 | 0.00 | 0.00 | — | — |
| 33.7 | [The Last Remote Sensing Model Built by Humans? Casting Any Earth Observation Task as Recursive Self-Improvement](https://arxiv.org/abs/2610.18845) `2610.18845` | 0.50 | 0.45 | 0.40 | 0.00 | 0.00 | — | — |
| 33.7 | [CREATE: Creativity Evaluation of VLMs using the Visual Remote Associates Test](https://arxiv.org/abs/2610.18569) `2610.18569` | 0.50 | 0.45 | 0.40 | 0.00 | 0.00 | — | — |
| 33.7 | [Empathy Is Not the Scapegoat for Declines in Factuality and Reliability](https://arxiv.org/abs/2610.18541) `2610.18541` | 0.50 | 0.45 | 0.40 | 0.00 | 0.00 | — | — |
| 33.7 | [Aligning Faithfulness for Theory-Scale Autoformalization](https://arxiv.org/abs/2610.18498) `2610.18498` | 0.50 | 0.45 | 0.40 | 0.00 | 0.00 | — | — |
| 33.7 | [Reliable Memory from Unstable Growth: Curating Self-Written Agent Playbooks with Randomized Experiments](https://arxiv.org/abs/2610.18427) `2610.18427` | 0.50 | 0.45 | 0.40 | 0.00 | 0.00 | — | — |
| 33.7 | [The Scored Population, the Operating Point, and the Pipeline: Three Objects That Decide What an Open-Web Event System Looks Like](https://arxiv.org/abs/2610.17814) `2610.17814` | 0.50 | 0.40 | 0.30 | 0.00 | 0.20 | — | — |
| 33.7 | [Can Rényi Accountants Beat Lossless Accountants for DP-SGD?](https://arxiv.org/abs/2610.17120) `2610.17120` | 0.50 | 0.40 | 0.45 | 0.00 | 0.00 | — | — |
| 32.6 | [Scaling Online Optimal Transport Minibatches for Flow Matching](https://arxiv.org/abs/2610.18838) `2610.18838` | 0.50 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 32.6 | [SP$^2$ec: Adaptive Self-Speculative Decoding for Vision-Language Models](https://arxiv.org/abs/2610.18735) `2610.18735` | 0.30 | 0.40 | 0.55 | 0.00 | 0.20 | — | — |
| 32.6 | [Bayesian Test-Time Inference of Task-Aligned Similarity from Weak Interactive Feedback](https://arxiv.org/abs/2610.18656) `2610.18656` | 0.50 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 32.6 | [Kwai Summary Attention V2 Technical Report](https://arxiv.org/abs/2610.18654) `2610.18654` | 0.50 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 32.6 | [Grid or Nodes? Probing Object Memory in World Models Under Partial Observability](https://arxiv.org/abs/2610.18637) `2610.18637` | 0.50 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 32.6 | [Imprinting: Knowledge Distillation Can Shift Student Behavior](https://arxiv.org/abs/2610.18517) `2610.18517` | 0.50 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 32.6 | [Readout Scaling Shapes Sparse Feature Recovery](https://arxiv.org/abs/2610.18462) `2610.18462` | 0.50 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 32.6 | [Teacher-Guided Tool Learning for Small Language Models via Bidirectional KL Constraints](https://arxiv.org/abs/2610.18459) `2610.18459` | 0.50 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 32.6 | [Tracing Money Flows in the Multi-Token Cryptocurrency Network](https://arxiv.org/abs/2610.16981) `2610.16981` | 0.50 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 32.6 | [Comparing GenAI to Machine Learning and Manual Review for eDiscovery](https://arxiv.org/abs/2610.16925) `2610.16925` | 0.50 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 32.6 | [From Volpiano to Pixels: 30,000 Lines of OMR Ground Truth for Medieval Chant](https://arxiv.org/abs/2610.16862) `2610.16862` | 0.50 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 32.6 | [Understanding Circulant Permutation to Extend the CMinHash Estimator](https://arxiv.org/abs/2610.18551) `2610.18551` | 0.50 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 32.6 | [Linear Regression Beyond Covariance: Latent Slope Spectra from Higher-Order Cumulants](https://arxiv.org/abs/2610.17772) `2610.17772` | 0.50 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 32.6 | [A Generalized McDiarmid Inequality: Sample Reuse and Non-Uniform Unbounded Differences](https://arxiv.org/abs/2610.17159) `2610.17159` | 0.50 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 32.6 | [Feature Normalization under Finite Training Budgets](https://arxiv.org/abs/2610.16845) `2610.16845` | 0.50 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 32.6 | [On the Calibration of Crowdsourcing Aggregators](https://arxiv.org/abs/2610.16815) `2610.16815` | 0.30 | 0.40 | 0.70 | 0.00 | 0.00 | — | — |
| 32.1 | [Data Free Training of Diffusion Models for Lattice Field Theories](https://arxiv.org/abs/2610.18863) `2610.18863` | 0.35 | 0.45 | 0.40 | 0.00 | 0.20 | — | — |
| 32.1 | [Fish Audio Drama 3 Technical Report](https://arxiv.org/abs/2610.18010) `2610.18010` | 0.00 | 0.40 | 0.40 | 1.00 | 0.30 | — | — |
| 32.1 | [DeltaTTG: Training Multi-Turn Agents with a Token-to-Go Progress Meter](https://arxiv.org/abs/2610.17666) `2610.17666` | 0.00 | 0.40 | 0.40 | 1.00 | 0.30 | — | — |
| 31.1 | [To Align or Not To Align: Check Your COMPASS Before You Train](https://arxiv.org/abs/2610.18638) `2610.18638` | 0.25 | 0.40 | 0.55 | 0.00 | 0.20 | — | — |
| 31.1 | [WorkRover: Scaling Long-Horizon Cowork Agent Training Via Recursive Environment Evolution](https://arxiv.org/abs/2610.18067) `2610.18067` | 0.00 | 0.45 | 0.30 | 1.00 | 0.30 | — | — |
| 30.9 | [DoubtLess: Training-Free Contrastive Decoding for Efficient Reasoning](https://arxiv.org/abs/2610.18862) `2610.18862` | 0.00 | 0.45 | 0.40 | 1.00 | 0.16 | — | — |
| 30.5 | [What Matters in Quality-Aware VLA Pretraining: A Benchmark and Empirical Study](https://arxiv.org/abs/2610.18869) `2610.18869` | 0.40 | 0.45 | 0.40 | 0.00 | 0.00 | — | — |
| 30.5 | [SMI: Ontology-Grounded Semantic IDs for Hierarchy-Aware Medical Concept Representation](https://arxiv.org/abs/2610.18672) `2610.18672` | 0.50 | 0.40 | 0.30 | 0.00 | 0.00 | — | — |
| 30.5 | [Breakdown of Conversational Involvement: A Joint Multi-Modal Analysis of Conversation Floors and F-formation Detection with Multiple Features](https://arxiv.org/abs/2610.18627) `2610.18627` | 0.50 | 0.40 | 0.30 | 0.00 | 0.00 | — | — |
| 30.5 | [SSEmb: Training a Generalizable Speech Retrieval Model Using Only Automatic Speech Recognition Data](https://arxiv.org/abs/2610.17813) `2610.17813` | 0.40 | 0.40 | 0.30 | 0.00 | 0.20 | — | — |
| 30.5 | [On the Statistical Mechanics of Attention](https://arxiv.org/abs/2610.17347) `2610.17347` | 0.50 | 0.40 | 0.30 | 0.00 | 0.00 | — | — |
| 30.5 | [GeoFix: Geometry Fixing for 4D Consistent Generation](https://arxiv.org/abs/2610.14293) `2610.14293` | 0.35 | 0.40 | 0.30 | 0.00 | 0.30 | — | — |
| 29.5 | [RALI@iKAT 2026: Retrieval-aware User Profile Denoising via Personalized Conversational Dense Retrieval](https://arxiv.org/abs/2610.17673) `2610.17673` | 0.30 | 0.40 | 0.40 | 0.00 | 0.20 | — | — |
| 28.9 | [LongR: Unleashing Long-Context Reasoning via Reinforcement Learning with Dense Utility Rewards](https://arxiv.org/abs/2602.05758) `2602.05758` | 0.30 | 0.40 | 0.30 | 0.00 | 0.30 | — | — |
| 28.4 | [Web$hell: A Unified Interface for Agentic Web Access](https://arxiv.org/abs/2610.13416) `2610.13416` | 0.30 | 0.45 | 0.30 | 0.00 | 0.20 | — | — |
| 27.4 | [Depth Memory](https://arxiv.org/abs/2610.18809) `2610.18809` | 0.50 | 0.40 | 0.00 | 0.00 | 0.20 | — | — |
| 27.4 | [Just Respect the Semantics: Image-Token Compression via Pre-LLM Semantic Variable-Resolution Tokenization and In-LLM Relevance Pruning](https://arxiv.org/abs/2610.18690) `2610.18690` | 0.30 | 0.45 | 0.40 | 0.00 | 0.00 | — | — |
| 27.4 | [Does self-distillation improve reasoning?](https://arxiv.org/abs/2610.18613) `2610.18613` | 0.00 | 0.40 | 0.40 | 1.00 | 0.00 | — | — |
| 27.4 | [Rethinking Projector Training in Multimodal LLMs](https://arxiv.org/abs/2610.18598) `2610.18598` | 0.30 | 0.45 | 0.40 | 0.00 | 0.00 | — | — |
| 27.4 | [Judges Follow the Reference: The Derivation Gap in Verifying Tool-Call Arguments](https://arxiv.org/abs/2610.18497) `2610.18497` | 0.30 | 0.45 | 0.40 | 0.00 | 0.00 | — | — |
| 27.4 | [Deeply unrolled alternating projections for phase retrieval](https://arxiv.org/abs/2610.17316) `2610.17316` | 0.50 | 0.40 | 0.00 | 0.00 | 0.20 | — | — |
| 26.3 | [Can an Agent Learn to Predict Its Own Failure? Outcome-Grounded Confidence Calibration in LLM Agents](https://arxiv.org/abs/2610.18614) `2610.18614` | 0.30 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 26.3 | [Where Do LLM Ensemble Gains Come From? Decomposing Candidate Value, Diversity, and Aggregation](https://arxiv.org/abs/2610.18786) `2610.18786` | 0.30 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 26.3 | [Looking Complete Is Not Being Grounded: Evaluating LLM Sitreps Against Human-Authored Situation Reports](https://arxiv.org/abs/2610.18762) `2610.18762` | 0.30 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 26.3 | [From Hindsight to Foresight: Dissecting Clinical Evidence Acquisition in Large Language Models](https://arxiv.org/abs/2610.18605) `2610.18605` | 0.30 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 26.3 | [Decomposing and Predicting Annotation Disagreement in Harmful Content Across Languages](https://arxiv.org/abs/2610.18588) `2610.18588` | 0.30 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 26.3 | [How Reliable Are System-One Decisions in Medicine? A Comparative Study of Jev, Jev-like Models, and LLMs](https://arxiv.org/abs/2610.18529) `2610.18529` | 0.30 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 26.3 | [Copying Is Not Grounding: Training Faithful Long-Form RAG with Authored-Sentence Credit](https://arxiv.org/abs/2610.18442) `2610.18442` | 0.30 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 26.3 | [Fresh at Decision, Stale at Action: Routing New Information to LLM Decisions Already Made](https://arxiv.org/abs/2610.18126) `2610.18126` | 0.30 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 26.3 | [Separating Prompt Harmfulness from Interaction Safety via Partial Generation](https://arxiv.org/abs/2610.16761) `2610.16761` | 0.30 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 26.3 | [Learning to Detect Shifts in Unstructured Data Streams via Monitoring-by-Betting](https://arxiv.org/abs/2610.17399) `2610.17399` | 0.30 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 25.8 | [UniW(A)M: A Unified Mixture-of-Experts Framework for World and World-Action Models](https://arxiv.org/abs/2610.18697) `2610.18697` | 0.35 | 0.40 | 0.30 | 0.00 | 0.00 | — | — |
| 25.3 | [Beyond Answer Voting: Fine-Grained Correction of LLM Reasoning through Graph Alignment across Sampled Traces](https://arxiv.org/abs/2610.18606) `2610.18606` | 0.30 | 0.45 | 0.30 | 0.00 | 0.00 | — | — |
| 25.3 | [Operator Surrogates](https://arxiv.org/abs/2610.18210) `2610.18210` | 0.50 | 0.40 | 0.05 | 0.00 | 0.00 | — | — |
| 24.7 | [Does Pretraining Teach Chemical Language Models Chemistry?](https://arxiv.org/abs/2610.18651) `2610.18651` | 0.25 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 24.7 | [The Geometry of Language Model Logits: Generic Structure and Learned Departures](https://arxiv.org/abs/2610.18520) `2610.18520` | 0.25 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 24.2 | [MFG: Multi-turn Feedback Guided Molecular Optimization](https://arxiv.org/abs/2610.18851) `2610.18851` | 0.30 | 0.40 | 0.30 | 0.00 | 0.00 | — | — |
| 24.2 | [Registration is a Representation Problem: Solving Whole-Body PET/CT Registration](https://arxiv.org/abs/2610.18804) `2610.18804` | 0.30 | 0.40 | 0.30 | 0.00 | 0.00 | — | — |
| 24.2 | [Emergent Loop Geometry in Gauge-Equivariant Graph Neural Networks](https://arxiv.org/abs/2610.18742) `2610.18742` | 0.50 | 0.40 | 0.00 | 0.00 | 0.00 | — | — |
| 24.2 | [AIRRI: Towards a Human-Readable, Machine-Resistant Reading Interface](https://arxiv.org/abs/2610.18724) `2610.18724` | 0.30 | 0.40 | 0.30 | 0.00 | 0.00 | — | — |
| 24.2 | [ChunkAct: Action Chunking for Agentic Robotics](https://arxiv.org/abs/2610.18874) `2610.18874` | 0.30 | 0.40 | 0.30 | 0.00 | 0.00 | — | — |
| 24.2 | [BLADE: Auditing Information Use for Failure Attribution in LLM Multi-Agent Systems](https://arxiv.org/abs/2610.18780) `2610.18780` | 0.30 | 0.40 | 0.30 | 0.00 | 0.00 | — | — |
| 23.2 | [Code-Switching Detection via Language-Selective Dimensions in Speech Foundation Models](https://arxiv.org/abs/2610.18726) `2610.18726` | 0.20 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 23.2 | [The Last Mile of Coding Agents for Long-Document Tasks: An Empirical Study on Clinical Trial Protocol Digitization](https://arxiv.org/abs/2610.18507) `2610.18507` | 0.50 | 0.05 | 0.30 | 0.00 | 0.00 | — | — |
| 22.6 | [MajorTOM:Elliot Pretrain, A Multi-Sensor, Multi-Temporal Pretraining Dataset for Earth Observation](https://arxiv.org/abs/2610.18624) `2610.18624` | 0.05 | 0.45 | 0.40 | 0.00 | 0.20 | — | — |
| 21.6 | [Two For The Price Of One: Curvature-Aware Pruning Preserves Influence Functions Too](https://arxiv.org/abs/2610.18516) `2610.18516` | 0.05 | 0.40 | 0.40 | 0.00 | 0.20 | — | — |
| 21.6 | [Tiny Recursive Message Passing](https://arxiv.org/abs/2610.18649) `2610.18649` | 0.15 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 21.1 | [From Relation-Specific Heads to Shared Interventions: Contextual Entrainment in Language Models](https://arxiv.org/abs/2610.18674) `2610.18674` | 0.30 | 0.40 | 0.15 | 0.00 | 0.00 | — | — |
| 20.0 | [textGrain watermarking is harder to detect: can sum-tests still do it optimally?](https://arxiv.org/abs/2610.17655) `2610.17655` | 0.00 | 0.40 | 0.40 | 0.00 | 0.20 | — | — |
| 19.5 | [Tev: Scaling Long-Horizon Agentic Environments with Dense Progress Rewards](https://arxiv.org/abs/2610.18868) `2610.18868` | 0.00 | 0.40 | 0.30 | 0.00 | 0.30 | — | — |
| 19.5 | [Model-Based Online Decision Making via Generative Trajectory Planning](https://arxiv.org/abs/2610.18677) `2610.18677` | 0.35 | 0.40 | 0.00 | 0.00 | 0.00 | — | — |
| 18.9 | [OmniCompass: A Holistic Benchmark for Omni-Modal Video Grounding and Understanding](https://arxiv.org/abs/2610.18829) `2610.18829` | 0.30 | 0.05 | 0.40 | 0.00 | 0.00 | — | — |
| 17.9 | [How Low Can You Go: Fine-Tuning LLMs with Bayesian Optimization (LlMBO)](https://arxiv.org/abs/2610.18504) `2610.18504` | 0.00 | 0.45 | 0.40 | 0.00 | 0.00 | — | — |
| 17.9 | [The Many Morals of AI: The cautionary bias of North American news narratives in global perspective](https://arxiv.org/abs/2610.18573) `2610.18573` | 0.30 | 0.40 | 0.00 | 0.00 | 0.00 | — | — |
| 16.8 | [Learning When to Transfer: Reinforcement Learning for Carbon-Footprint Modeling in New Manufacturing Domains](https://arxiv.org/abs/2610.18730) `2610.18730` | 0.00 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 16.8 | [PreciseCoder: Faithful and Generalizable Debugging Beyond Unit-Test Rewards](https://arxiv.org/abs/2610.18755) `2610.18755` | 0.00 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 16.8 | [More Reader Data or a Better Teacher? Updating a Model of Crowd Highlights in Late 2026](https://arxiv.org/abs/2610.17768) `2610.17768` | 0.00 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 13.2 | [Geometric Calibration of Supervised Fine-Tuning](https://arxiv.org/abs/2610.18634) `2610.18634` | 0.15 | 0.40 | 0.00 | 0.00 | 0.00 | — | — |
| 12.6 | [Web Scale on a Budget: Tail-Aware Retrieval Evaluation](https://arxiv.org/abs/2610.17986) `2610.17986` | 0.00 | 0.05 | 0.40 | 0.00 | 0.20 | — | — |
| 10.0 | [Back-Translation Still Matters: Leveraging Monolingual Data for Low-Resource Machine Translation with LLMs](https://arxiv.org/abs/2610.18818) `2610.18818` | 0.05 | 0.40 | 0.00 | 0.00 | 0.00 | — | — |

Scores: weighted mean of rubric criteria (0–1 each) × 100; see README for the rubric.