# paper-forge daily report — 2026-10-07

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

## Reproductions

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

## Top candidates

### [Variance-Optimal Off-Policy Evaluation with Conjunct Effect Modeling](https://arxiv.org/abs/2610.08677)
`2610.08677` · score 62.6 · analysis: heuristic · feasibility 8/10

Off-policy evaluation (OPE) for contextual bandit policies becomes challenging when action-level importance weighting incurs excessive variance. Doubly robust (DR) estimation remains unbiased under common support but retains these high-variance action-level weights.

Rubric evidence: + cheap/closed-form ("closed form"); + toy/synthetic experiments ("synthetic"); + small/synthetic data ("synthetic"); + public benchmarks ("benchmarks"); + concrete method ("closed form"); + formal statement ("theorem"); no public code found; + keywords: bandit

### [Structure, Not Belief: Correlated Thompson Sampling from LLM-Derived Covariance in Combinatorial Semi-Bandits](https://arxiv.org/abs/2610.07470)
`2610.07470` · score 61.6 · analysis: heuristic · feasibility 8/10

Combinatorial Thompson sampling (CTS) draws independent posterior samples for every arm, so its exploration dynamics ignore any relation among arms. We study a minimal change to those dynamics: an LLM is queried once for a partition of the arms, the partition becomes a positive-definite correlation matrix $Σ$ through an RBF kernel on cluster ranks, and the per-round posterior sample is drawn with covariance $Σ$ while the Beta posteriors are updated from real rewards only, so the LLM shapes how the sampler moves, not what it believes.

Rubric evidence: - LLM-based ("LLM"); + toy/synthetic experiments ("synthetic"); + small/synthetic data ("synthetic"); + public benchmarks ("benchmark"); + explicit algorithm/pseudocode ("pseudocode"); + formal statement ("bound"); + claims simplicity ("simple"); no public code found; + keywords: sampling, kernel, bandit

### [Linear Bandits under Exact Sliding-Window Constraints](https://arxiv.org/abs/2610.08745)
`2610.08745` · score 58.4 · analysis: heuristic · feasibility 8/10

We study linear bandits under exact sliding-window constraints, where every consecutive block of actions must belong to a prescribed feasible set. In the offline setting, where the reward function is known, we show that convexity and cyclic-shift invariance make a stationary solution optimal when $w\mid T$ and within an additive $O(w)$ gap otherwise.

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

### [CNet: A Complex-Valued Deep Learning Framework with Wirtinger Autodifferentiation and FFT--Hadamard Convolution](https://arxiv.org/abs/2610.08592)
`2610.08592` · score 56.3 · analysis: heuristic · feasibility 5/10

CNet is a C++/CUDA framework for building and training deep complex-valued neural networks (CVNNs) and, more generally, for optimizing complex-valued functions by gradient descent with Wirtinger (CR-calculus) derivatives. It takes a physics-native stance: a network is a cascade of complex -- and often unitary (the DFT) -- operations acting on an amplitude vector, and classification is a Born-rule measurement $p_k = |z_k|^2 / \|z\|^2$ rather than a softmax over real logits.

Rubric evidence: + CPU-friendly ("CPU"); + concrete method ("closed form"); - system/framework (many moving parts) ("Framework"); + code: https://github.com/crasmarum/CNet; + keywords: optimizer, kernel

### [Beyond Marginal Monitoring: Distributed Joint-Distribution Testing for Data Concept Drift in Large Scale E-Commerce Operations](https://arxiv.org/abs/2610.08132)
`2610.08132` · score 55.3 · analysis: heuristic · feasibility 6/10

Concept drift threatens production machine learning, yet the empirical behavior of multivariate two-sample drift detectors at scale remains under-characterized. Existing benchmarks rarely address the hundreds of millions of rows and high-cardinality features typical of industrial-operational datasets.

Rubric evidence: + toy/synthetic experiments ("synthetic"); + small/synthetic data ("synthetic"); + public benchmarks ("benchmarks"); no public code found; + keywords: sampling, kernel, ranking

### [Accent Analogy Guidance: More Speaker Similarity at Equal Accent in Cross-Lingual Voice Cloning](https://arxiv.org/abs/2609.29123)
`2609.29123` · score 52.6 · analysis: heuristic · feasibility 5/10

In cross-lingual zero-shot text-to-speech, the accent of the reference leaks into the target speech. We propose accent analogy guidance (AAG), a training-free sampler term that subtracts an accent direction estimated from the model's own predictions for one synthetic voice rendered in both languages, so the voice cancels and only the accent remains.

Rubric evidence: - LLM-based ("LLM"); + toy/synthetic experiments ("synthetic"); + small/synthetic data ("synthetic"); + code: https://github.com/yoomee-cho/accent-analogy-guidance; + 4 HF upvotes

### [Confidence-Ordering Reversal under Contextual Priors in Neural Decoding](https://arxiv.org/abs/2610.08229)
`2610.08229` · score 52.6 · analysis: heuristic · feasibility 5/10

Contextual priors improve neural-to-language decoding by reshaping candidate scores. However, confidence is read from the same reshaped scores, so the errors a prior leaves behind can become more confident with no change in accuracy to reveal it.

Rubric evidence: + formal statement ("bound"); + code: https://github.com/AmadeusFake/NeuDecodingConfReversal; + keywords: retrieval, ranking

### [Token-Budgeted Escalation for Financial Document QA: Cost Is Predictable, Benefit Is the Bottleneck](https://arxiv.org/abs/2610.07760)
`2610.07760` · score 52.6 · analysis: heuristic · feasibility 5/10

Retrieval-augmented generation systems can route difficult queries to deeper context, but batch deployments must allocate a shared token budget across calls whose costs vary by query. We formulate selective escalation as finite-batch allocation for financial document question answering.

Rubric evidence: + formal statement ("bound"); + code: https://github.com/junru-zhu/token-budgeted-escalation-financial-qa; + keywords: retrieval, ranking

### [Conditional Flow Matching for Transport Between Markov Processes](https://arxiv.org/abs/2610.07229)
`2610.07229` · score 51.1 · analysis: heuristic · feasibility 6/10

Motivated by sequence-to-sequence transport in the context time-series domain adaptation, we study the problem of transportation between trajectories of Markov processes. Given a limited number of trajectories from source distribution and the target distribution, we formulate a flow matching based algorithm which learns a transport map from the source to target trajectory distribution, while preserving the Markov structure.

Claimed: accuracy 43.87% (abstract (regex extraction, unverified))
Rubric evidence: + toy/synthetic experiments ("synthetic"); + small/synthetic data ("synthetic"); + formal statement ("bound"); no public code found; + keywords: retrieval

### [From Evidence to Action: How Tool-Using Agents Fail](https://arxiv.org/abs/2610.07753)
`2610.07753` · score 51.1 · analysis: heuristic · feasibility 5/10

Tool-using agents make consequential changes to external state, yet correct outcomes do not guarantee that their actions were supported by evidence established beforehand. We study where this evidence-to-action chain breaks as agents move from deciding whether to act to executing single actions and dependent workflows.

Rubric evidence: + formal statement ("bound"); + code: https://github.com/caoshidong66/safeact; + 30 HF upvotes

## All screened papers

| score | paper | compute | data | clarity | code | interest | analysis | laptop-feasible |
|---:|---|---:|---:|---:|---:|---:|---|---|
| 62.6 | [Variance-Optimal Off-Policy Evaluation with Conjunct Effect Modeling](https://arxiv.org/abs/2610.08677) `2610.08677` | 0.85 | 0.80 | 0.75 | 0.00 | 0.20 | heuristic | True |
| 61.6 | [Structure, Not Belief: Correlated Thompson Sampling from LLM-Derived Covariance in Combinatorial Semi-Bandits](https://arxiv.org/abs/2610.07470) `2610.07470` | 0.45 | 0.80 | 1.00 | 0.00 | 0.60 | heuristic | False |
| 58.4 | [Linear Bandits under Exact Sliding-Window Constraints](https://arxiv.org/abs/2610.08745) `2610.08745` | 0.65 | 0.80 | 0.85 | 0.00 | 0.20 | heuristic | True |
| 56.3 | [CNet: A Complex-Valued Deep Learning Framework with Wirtinger Autodifferentiation and FFT--Hadamard Convolution](https://arxiv.org/abs/2610.08592) `2610.08592` | 0.65 | 0.40 | 0.50 | 1.00 | 0.40 | heuristic | True |
| 55.3 | [Beyond Marginal Monitoring: Distributed Joint-Distribution Testing for Data Concept Drift in Large Scale E-Commerce Operations](https://arxiv.org/abs/2610.08132) `2610.08132` | 0.65 | 0.80 | 0.40 | 0.00 | 0.60 | heuristic | True |
| 52.6 | [Accent Analogy Guidance: More Speaker Similarity at Equal Accent in Cross-Lingual Voice Cloning](https://arxiv.org/abs/2609.29123) `2609.29123` | 0.45 | 0.75 | 0.40 | 1.00 | 0.23 | heuristic | False |
| 52.6 | [Confidence-Ordering Reversal under Contextual Priors in Neural Decoding](https://arxiv.org/abs/2610.08229) `2610.08229` | 0.50 | 0.40 | 0.55 | 1.00 | 0.40 | heuristic | True |
| 52.6 | [Token-Budgeted Escalation for Financial Document QA: Cost Is Predictable, Benefit Is the Bottleneck](https://arxiv.org/abs/2610.07760) `2610.07760` | 0.50 | 0.40 | 0.55 | 1.00 | 0.40 | heuristic | True |
| 51.1 | [Conditional Flow Matching for Transport Between Markov Processes](https://arxiv.org/abs/2610.07229) `2610.07229` | 0.65 | 0.75 | 0.55 | 0.00 | 0.20 | heuristic | True |
| 51.1 | [From Evidence to Action: How Tool-Using Agents Fail](https://arxiv.org/abs/2610.07753) `2610.07753` | 0.50 | 0.40 | 0.55 | 1.00 | 0.30 | heuristic | True |
| 49.5 | [RAGFlip: Measuring Query-Level Negative Flips in Retriever Upgrades](https://arxiv.org/abs/2610.07266) `2610.07266` | 0.50 | 0.40 | 0.55 | 1.00 | 0.20 | — | — |
| 48.4 | [The Right Memory in the Wrong Context: Verifying Retrieval Admissibility in Long-Term Agent Memory](https://arxiv.org/abs/2610.07309) `2610.07309` | 0.50 | 0.45 | 0.30 | 1.00 | 0.40 | — | — |
| 47.9 | [How Many Independent Samples Does a Satellite Image Contain? Generalization Bounds for Spatially Dependent Data](https://arxiv.org/abs/2610.08227) `2610.08227` | 0.65 | 0.75 | 0.55 | 0.00 | 0.00 | — | — |
| 47.9 | [HyperNSDE: Personalized Neural SDEs for Joint Static-Longitudinal Clinical Data Generation](https://arxiv.org/abs/2610.07383) `2610.07383` | 0.65 | 0.75 | 0.40 | 0.00 | 0.20 | — | — |
| 46.8 | [Sherpa: Teaching LLMs to Teach Adaptively](https://arxiv.org/abs/2610.08778) `2610.08778` | 0.45 | 0.75 | 0.30 | 1.00 | 0.00 | — | — |
| 46.8 | [Spectra: Exact Component Transport for Test-Time Prior Adaptation in Simulation-Based Inference](https://arxiv.org/abs/2610.08021) `2610.08021` | 0.45 | 0.80 | 0.60 | 0.00 | 0.20 | — | — |
| 46.8 | [Selection-Based Structured Reasoning: Toward Efficient Multimodal Search Agents](https://arxiv.org/abs/2610.01892) `2610.01892` | 0.50 | 0.45 | 0.30 | 1.00 | 0.30 | — | — |
| 46.8 | [AdvSim2Real : Training Web Agents Against Adaptive Prompt Injection in a Web World Model](https://arxiv.org/abs/2610.08773) `2610.08773` | 0.50 | 0.40 | 0.40 | 1.00 | 0.23 | — | — |
| 46.7 | [Judged Useless, Queried Anyway: Tool-Using Agents Rarely Turn Their Own Evidence Judgments into Stopping Decisions](https://arxiv.org/abs/2610.06191) `2610.06191` | 0.50 | 0.40 | 0.30 | 1.00 | 0.36 | — | — |
| 46.3 | [Steering Diffusion Models to Rare Events with Sequential Monte Carlo](https://arxiv.org/abs/2610.08652) `2610.08652` | 0.50 | 0.75 | 0.40 | 0.00 | 0.40 | — | — |
| 46.3 | [Early Memory Selection for Balanced Adam](https://arxiv.org/abs/2610.08624) `2610.08624` | 0.50 | 0.40 | 0.40 | 1.00 | 0.20 | — | — |
| 46.3 | [Beyond Perturbation Magnitude: Direction-Dependent Responses in Multimodal Geometric Representations](https://arxiv.org/abs/2610.08533) `2610.08533` | 0.70 | 0.40 | 0.60 | 0.00 | 0.20 | — | — |
| 46.3 | [Beyond Successor Accuracy: State Retention for Recursive Self-Improvement in Recommendation](https://arxiv.org/abs/2610.07105) `2610.07105` | 0.50 | 0.40 | 0.40 | 1.00 | 0.20 | — | — |
| 46.3 | [Redundancy and synergy in multivariate Gaussians via the Blackwell order](https://arxiv.org/abs/2610.07360) `2610.07360` | 0.70 | 0.40 | 0.75 | 0.00 | 0.00 | — | — |
| 45.8 | [The Failure Is in the Readout: Fine-Grained Emotion Recognition Benchmarks Measure Elicitation, Not Perception](https://arxiv.org/abs/2610.08162) `2610.08162` | 0.65 | 0.80 | 0.40 | 0.00 | 0.00 | — | — |
| 45.3 | [Agentic AutoRAG: RAG Pipeline Optimization through Reasoning-Driven Agents](https://arxiv.org/abs/2610.08452) `2610.08452` | 0.30 | 0.45 | 0.30 | 1.00 | 0.60 | — | — |
| 44.7 | [DeepAJM: Deep Association Joint Model for Irregularly Sampled data](https://arxiv.org/abs/2610.07388) `2610.07388` | 0.65 | 0.75 | 0.40 | 0.00 | 0.00 | — | — |
| 44.7 | [MEND: RL For Flow Models via Proximal Velocity Matching](https://arxiv.org/abs/2610.05954) `2610.05954` | 0.50 | 0.40 | 0.40 | 1.00 | 0.10 | — | — |
| 44.7 | [Source Identification Is Not Fitness Testing: Measuring the Limits of Synthetic-Data Attribution](https://arxiv.org/abs/2610.00417) `2610.00417` | 0.65 | 0.75 | 0.40 | 0.00 | 0.00 | — | — |
| 44.2 | [Rethinking Semantic ID Construction for Generative Recommendation: SimHash with Parallel Decoding and Semantic Alignment](https://arxiv.org/abs/2610.07402) `2610.07402` | 0.50 | 0.40 | 0.45 | 1.00 | 0.00 | — | — |
| 43.7 | [Nash Social Welfare for Multi Armed Bandits: Trajectory-wise Expected and High Probability Regret](https://arxiv.org/abs/2610.07737) `2610.07737` | 0.65 | 0.40 | 0.55 | 0.00 | 0.20 | — | — |
| 43.7 | [Two-Sample Testing for Random Graphs without Vertex Correspondence](https://arxiv.org/abs/2610.07503) `2610.07503` | 0.65 | 0.40 | 0.70 | 0.00 | 0.00 | — | — |
| 43.2 | [MemFLoRA: Memory-Floor LoRA for CNN Adaptation at the Edge](https://arxiv.org/abs/2610.08669) `2610.08669` | 0.50 | 0.40 | 0.40 | 1.00 | 0.00 | — | — |
| 43.2 | [Making COMET Comparable Across Scripts: Diagnosis and Correction of Tokeniser-Induced Script Bias in Indic MT Evaluation](https://arxiv.org/abs/2610.08159) `2610.08159` | 0.50 | 0.40 | 0.40 | 1.00 | 0.00 | — | — |
| 43.2 | [Detecting a Shift Is Not Enough: Exact Minimax Limits of Linear Representation Repair](https://arxiv.org/abs/2610.08069) `2610.08069` | 0.50 | 0.40 | 0.40 | 1.00 | 0.00 | — | — |
| 43.0 | [DAEDALUS: Bootstrapping Agent Memory from Self-Generated Tasks](https://arxiv.org/abs/2610.08048) `2610.08048` | 0.30 | 0.45 | 0.30 | 1.00 | 0.46 | — | — |
| 42.6 | [Probabilistic Counterfactual Inference for Discrete Outcomes in Gaussian-Process Causal Models](https://arxiv.org/abs/2610.08689) `2610.08689` | 0.65 | 0.75 | 0.30 | 0.00 | 0.00 | — | — |
| 42.6 | [Empirical-Bayes spectral partial pooling across related tasks](https://arxiv.org/abs/2610.07284) `2610.07284` | 0.65 | 0.75 | 0.30 | 0.00 | 0.00 | — | — |
| 42.1 | [Reinforcement Learning for Hierarchical Reasoning Rewards: Minimax-Optimal Rates with Transformers](https://arxiv.org/abs/2610.08561) `2610.08561` | 0.50 | 0.40 | 0.55 | 0.00 | 0.40 | — | — |
| 42.1 | [CoDe-LoRA: Mitigating the Orthogonality Dilemma in Continual Learning of LLMs via Knowledge Consolidation and Decoupling](https://arxiv.org/abs/2610.08312) `2610.08312` | 0.30 | 0.45 | 0.60 | 1.00 | 0.00 | — | — |
| 42.1 | [High-dimensional online calibration from harmonic weights](https://arxiv.org/abs/2610.07740) `2610.07740` | 0.70 | 0.40 | 0.55 | 0.00 | 0.00 | — | — |
| 41.6 | [ProximalFM: Amortized Proximal Causal Inference under Hidden Confounding](https://arxiv.org/abs/2610.08078) `2610.08078` | 0.45 | 0.75 | 0.40 | 0.00 | 0.20 | — | — |
| 41.6 | [Learning Discriminative Geometry for Drifting Models](https://arxiv.org/abs/2610.04703) `2610.04703` | 0.25 | 0.40 | 0.40 | 1.00 | 0.40 | — | — |
| 41.1 | [MetaLearnNCA: Few-Shot Offline Meta-Learning via Interacting Neural Cellular Automata](https://arxiv.org/abs/2610.08479) `2610.08479` | 0.50 | 0.75 | 0.30 | 0.00 | 0.20 | — | — |
| 41.1 | [One-Shot Private Confidence Regions via Resampling](https://arxiv.org/abs/2610.08460) `2610.08460` | 0.50 | 0.40 | 0.65 | 0.00 | 0.20 | — | — |
| 40.5 | [When does conformal calibration need censoring weights? Cause-of-failure prediction sets under competing risks](https://arxiv.org/abs/2610.08602) `2610.08602` | 0.65 | 0.40 | 0.55 | 0.00 | 0.00 | — | — |
| 40.5 | [Vine Copula VAR:From Recursive Margins to Joint Forecast Inference](https://arxiv.org/abs/2610.07589) `2610.07589` | 0.65 | 0.40 | 0.40 | 0.00 | 0.20 | — | — |
| 40.5 | [DiVeR: Decision-Critical Verifier Learning for VLA Test-Time Scaling](https://arxiv.org/abs/2610.04933) `2610.04933` | 0.50 | 0.40 | 0.40 | 0.00 | 0.50 | — | — |
| 40.0 | [Random Feature Gaussian Process Attention: Linear-Time Probabilistic Attention with Calibrated Uncertainty](https://arxiv.org/abs/2610.08578) `2610.08578` | 0.70 | 0.40 | 0.30 | 0.00 | 0.20 | — | — |
| 40.0 | [WildMatch: Weakly Supervised Image Matcher Adaptation for Wildlife Re-Identification](https://arxiv.org/abs/2610.07384) `2610.07384` | 0.25 | 0.40 | 0.40 | 1.00 | 0.30 | — | — |
| 40.0 | [Bayesian Optimization on Function Spaces via Sparse RKHS Manifolds](https://arxiv.org/abs/2610.07417) `2610.07417` | 0.50 | 0.45 | 0.55 | 0.00 | 0.20 | — | — |
| 39.5 | [Knowing When Not to Answer: Cross-Domain and Multi-Turn Generalization of Latent Underspecification Signals](https://arxiv.org/abs/2610.08413) `2610.08413` | 0.45 | 0.80 | 0.40 | 0.00 | 0.00 | — | — |
| 39.5 | [Making LLMs Say What They Think: Measuring and Improving CoT-Interpretability Alignment](https://arxiv.org/abs/2609.38972) `2609.38972` | 0.30 | 0.40 | 0.30 | 1.00 | 0.30 | — | — |
| 38.9 | [Singular Value Decomposition: A Geometric Rediscovery, Where Proofs Become Algorithms](https://arxiv.org/abs/2610.08565) `2610.08565` | 0.50 | 0.40 | 0.55 | 0.00 | 0.20 | — | — |
| 38.9 | [Foresight-over-Graph: Reasoning Beyond Local Horizons for Knowledge Base Question Answering](https://arxiv.org/abs/2610.08388) `2610.08388` | 0.30 | 0.45 | 0.30 | 1.00 | 0.20 | — | — |
| 38.9 | [Two-Sample Testing via Generative Processes](https://arxiv.org/abs/2610.08277) `2610.08277` | 0.50 | 0.40 | 0.55 | 0.00 | 0.20 | — | — |
| 38.9 | [How Inefficient Is Natural Gradient Descent? From Exact Optimality to Θ( \sqrt{ \log d } ) Divergence](https://arxiv.org/abs/2610.07228) `2610.07228` | 0.50 | 0.40 | 0.55 | 0.00 | 0.20 | — | — |
| 38.4 | [X-OPM: Explainable Automatic Digital On-Chip Power Modeling for Enhanced Robustness](https://arxiv.org/abs/2610.08502) `2610.08502` | 0.65 | 0.40 | 0.30 | 0.00 | 0.20 | — | — |
| 38.4 | [GeneICL: A Tabular Foundation Model for Bulk Transcriptomics](https://arxiv.org/abs/2610.08694) `2610.08694` | 0.35 | 0.75 | 0.55 | 0.00 | 0.00 | — | — |
| 38.4 | [Language Carries the Expert's Impression: Instrument-Anchored LLM Judges Transfer Counseling-Quality Assessment and Beat In-Domain Training](https://arxiv.org/abs/2610.08055) `2610.08055` | 0.45 | 0.75 | 0.40 | 0.00 | 0.00 | — | — |
| 37.9 | [Co-Evolving Paths and Flows via Path-Flow Alignment](https://arxiv.org/abs/2610.08717) `2610.08717` | 0.50 | 0.00 | 0.40 | 1.00 | 0.20 | — | — |
| 37.9 | [Multi-Label Perceptual Bug Detection in Video Games using Deep Learning on Gameplay Footage](https://arxiv.org/abs/2610.08593) `2610.08593` | 0.50 | 0.45 | 0.60 | 0.00 | 0.00 | — | — |
| 37.9 | [Learning to Retrieve via Reinforcement Learning in Embedding Space](https://arxiv.org/abs/2610.07731) `2610.07731` | 0.50 | 0.45 | 0.30 | 0.00 | 0.40 | — | — |
| 37.4 | [Information-Dense Synthesis for Molecular Discovery](https://arxiv.org/abs/2610.08495) `2610.08495` | 0.65 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 37.4 | [Rethinking Cross-Tokenizer On-Policy Distillation: From Alignment Coverage to Supervision Reliability](https://arxiv.org/abs/2610.08448) `2610.08448` | 0.50 | 0.40 | 0.40 | 0.00 | 0.30 | — | — |
| 37.4 | [Scalable Regularized Vector Multiplicative Error Models for Positive-valued Financial Time Series](https://arxiv.org/abs/2610.08443) `2610.08443` | 0.65 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 37.4 | [Anytime-valid simulation-based hypothesis testing](https://arxiv.org/abs/2610.08210) `2610.08210` | 0.65 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 37.4 | [Active Feature Acquisition for Cost-Efficient Temporal Prediction with Reduced Participant Burden](https://arxiv.org/abs/2610.07452) `2610.07452` | 0.65 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 37.4 | [Adaptive Latent Capacity for World Models](https://arxiv.org/abs/2609.32921) `2609.32921` | 0.50 | 0.40 | 0.40 | 0.00 | 0.30 | — | — |
| 36.8 | [When Forgetting is not Catastrophic: On the Mechanics of Spurious Forgetting](https://arxiv.org/abs/2610.08718) `2610.08718` | 0.40 | 0.75 | 0.40 | 0.00 | 0.00 | — | — |
| 36.8 | [A Systematic Study of Small Language Models on Abstract Reasoning Tasks](https://arxiv.org/abs/2610.08680) `2610.08680` | 0.50 | 0.45 | 0.40 | 0.00 | 0.20 | — | — |
| 36.8 | [RAG-PIBench: A Leakage-Aware Benchmark for Prompt-Injection Detection in Trustworthy RAG Systems](https://arxiv.org/abs/2610.08571) `2610.08571` | 0.50 | 0.45 | 0.40 | 0.00 | 0.20 | — | — |
| 36.8 | [Have I Seen Enough? Frozen Video-Language Models Encode Evidence Readiness](https://arxiv.org/abs/2610.08560) `2610.08560` | 0.50 | 0.45 | 0.40 | 0.00 | 0.20 | — | — |
| 36.8 | [Seeing the Context: Enhancing Recommender Systems with Image-Derived Contextual Signals](https://arxiv.org/abs/2610.08407) `2610.08407` | 0.50 | 0.40 | 0.60 | 0.00 | 0.00 | — | — |
| 36.3 | [Feature Information Dynamics in Diffusion](https://arxiv.org/abs/2610.08626) `2610.08626` | 0.35 | 0.40 | 0.30 | 1.00 | 0.00 | — | — |
| 35.8 | [Neural Petri flows for chemical reactions](https://arxiv.org/abs/2610.08750) `2610.08750` | 0.50 | 0.40 | 0.55 | 0.00 | 0.00 | — | — |
| 35.8 | [Reinforcement Learning with Conformal Action Sets: An Application to Sequential Recommendation](https://arxiv.org/abs/2610.08743) `2610.08743` | 0.50 | 0.40 | 0.55 | 0.00 | 0.00 | — | — |
| 35.8 | [On the Computational Tractability of Robust Bandits](https://arxiv.org/abs/2610.08740) `2610.08740` | 0.50 | 0.40 | 0.40 | 0.00 | 0.20 | — | — |
| 35.8 | [DeltaTTT: Layerwise Optimization for Nonlinear Recurrent Memory](https://arxiv.org/abs/2610.08553) `2610.08553` | 0.50 | 0.40 | 0.40 | 0.00 | 0.20 | — | — |
| 35.8 | [PHBA: Prefix-State Hybrid Block Attention](https://arxiv.org/abs/2610.08527) `2610.08527` | 0.50 | 0.40 | 0.40 | 0.00 | 0.20 | — | — |
| 35.8 | [Contrastive Learning for Aspect Representation towards Explainable Recommendation](https://arxiv.org/abs/2610.07761) `2610.07761` | 0.50 | 0.45 | 0.50 | 0.00 | 0.00 | — | — |
| 35.8 | [Classifications in modular restricted Boltzmann machines](https://arxiv.org/abs/2610.08612) `2610.08612` | 0.50 | 0.40 | 0.55 | 0.00 | 0.00 | — | — |
| 35.8 | [Symmetry-Aware Feature Learning: A Polynomial Separation for Multi-Index Models](https://arxiv.org/abs/2610.08420) `2610.08420` | 0.50 | 0.40 | 0.55 | 0.00 | 0.00 | — | — |
| 35.8 | [Network Intervention by Polling Strategic Agents](https://arxiv.org/abs/2610.08347) `2610.08347` | 0.50 | 0.40 | 0.40 | 0.00 | 0.20 | — | — |
| 35.8 | [Stochastic Gradient Descent Ascent is Suboptimal for Nonconvex-PL Min-Max Games](https://arxiv.org/abs/2610.07814) `2610.07814` | 0.50 | 0.40 | 0.55 | 0.00 | 0.00 | — | — |
| 35.8 | [Exact Calibration and Sharp Risk Geometry for Volume-Sampled Ridge Regression](https://arxiv.org/abs/2610.07721) `2610.07721` | 0.50 | 0.40 | 0.55 | 0.00 | 0.00 | — | — |
| 35.8 | [Explicit Asymptotic Bounds for Sequential Calibration Beyond $T^{2/3}$](https://arxiv.org/abs/2610.07623) `2610.07623` | 0.50 | 0.40 | 0.55 | 0.00 | 0.00 | — | — |
| 35.8 | [Is $\sqrt{d}$ Separation Necessary for Gradient EM to Learn Gaussian Mixtures in High Dimensions?](https://arxiv.org/abs/2610.07551) `2610.07551` | 0.50 | 0.40 | 0.55 | 0.00 | 0.00 | — | — |
| 35.3 | [High-Dimensional Statistical Inference for Sparse Support Vector Machines](https://arxiv.org/abs/2610.08345) `2610.08345` | 0.65 | 0.40 | 0.30 | 0.00 | 0.00 | — | — |
| 34.7 | [Align, Then Correct: Training-Free Two-Stage Low-Rank Compensation for Extremely Quantized Large Language Models](https://arxiv.org/abs/2610.08164) `2610.08164` | 0.50 | 0.40 | 0.50 | 0.00 | 0.00 | — | — |
| 34.7 | [Extending Pathwise Gradients to Discrete Random Variables via Finite-Order Relaxation](https://arxiv.org/abs/2610.07786) `2610.07786` | 0.50 | 0.45 | 0.45 | 0.00 | 0.00 | — | — |
| 34.2 | [Rapid Fredholm stabilization of the Kuramoto--Sivashinsky equation with unrestricted, spatially-varying anti-diffusion](https://arxiv.org/abs/2610.08764) `2610.08764` | 0.35 | 0.40 | 0.55 | 0.00 | 0.20 | — | — |
| 33.7 | [FlowCF: Sparse Counterfactual Explanations for Mixed-Type Tabular Data using Flow Matching](https://arxiv.org/abs/2610.08537) `2610.08537` | 0.50 | 0.45 | 0.40 | 0.00 | 0.00 | — | — |
| 33.7 | [A Systematic Study of Semantic ID Spaces for Generative Information Retrieval](https://arxiv.org/abs/2610.08732) `2610.08732` | 0.50 | 0.40 | 0.30 | 0.00 | 0.20 | — | — |
| 33.7 | [A Study of Prior Case Retrieval Using Lexical, Semantic, and Rhetorical Role Information in Indian Legal Documents](https://arxiv.org/abs/2610.07437) `2610.07437` | 0.30 | 0.40 | 0.60 | 0.00 | 0.20 | — | — |
| 33.7 | [Learning a Mixture of GFlowNets](https://arxiv.org/abs/2610.07562) `2610.07562` | 0.50 | 0.40 | 0.45 | 0.00 | 0.00 | — | — |
| 32.6 | [Denoising Hierarchical Representations: Joint Continuous Diffusion for Language Modeling](https://arxiv.org/abs/2610.08738) `2610.08738` | 0.10 | 0.45 | 0.45 | 1.00 | 0.00 | — | — |
| 32.6 | [Optimal and Efficient Online Inverse Optimization](https://arxiv.org/abs/2610.08735) `2610.08735` | 0.50 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 32.6 | [Does an Agent's History Tell You When Compaction Will Hurt? A Modest, Bounded Effect on the TRACE Paired-Replay Corpus](https://arxiv.org/abs/2610.08722) `2610.08722` | 0.50 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 32.6 | [Prediction-powered inference for time series across space](https://arxiv.org/abs/2610.08715) `2610.08715` | 0.50 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 32.6 | [FedDermaSeg: Federated Learning for Dermatological Image Segmentation](https://arxiv.org/abs/2610.08574) `2610.08574` | 0.50 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 32.6 | [Less Is More: A Leakage-Controlled Study of Dermoscopic Preprocessing for Joint Skin Lesion Classification and Segmentation with YOLO26](https://arxiv.org/abs/2610.08570) `2610.08570` | 0.50 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 32.6 | [Systemization of Knowledge (SoK): Human-Centered AI Safety for Youth](https://arxiv.org/abs/2610.08554) `2610.08554` | 0.50 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 32.6 | [AnyBottle: A Recipe to Only Keep the Concepts You Really Need](https://arxiv.org/abs/2610.08552) `2610.08552` | 0.50 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 32.6 | [Holdout Best-of-N: Unbiased Evaluation and Its Cost](https://arxiv.org/abs/2610.08719) `2610.08719` | 0.50 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 32.6 | [Generative AI translations in high-stakes emergency messaging](https://arxiv.org/abs/2610.08601) `2610.08601` | 0.50 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 32.6 | [How High Is 0.6? Floors, Ceilings, and Headroom in Interpretability Probing](https://arxiv.org/abs/2610.08544) `2610.08544` | 0.30 | 0.40 | 0.70 | 0.00 | 0.00 | — | — |
| 32.6 | [Language-model ratings of depression reflect the rater more than the patient](https://arxiv.org/abs/2610.08501) `2610.08501` | 0.50 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 32.6 | [STRUCTURALCOST: A controlled reading time dataset for modeling human sentence processing difficulty](https://arxiv.org/abs/2610.08208) `2610.08208` | 0.50 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 32.6 | [Adapting Generative Recommenders for Multi-Turn Interaction](https://arxiv.org/abs/2610.08136) `2610.08136` | 0.50 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 32.6 | [Spectral Recovery of Point Clouds from Noisy Geometric Graphs](https://arxiv.org/abs/2610.08634) `2610.08634` | 0.50 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 32.6 | [Stability of Measure-to-Measure Transformers on Sub-Gaussian Data](https://arxiv.org/abs/2610.07717) `2610.07717` | 0.50 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 32.6 | [Asymptotic Analysis of Empirical Risk Minimization on Entry-wise i.i.d. Heavy-Tailed Data](https://arxiv.org/abs/2610.07637) `2610.07637` | 0.50 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 32.6 | [The interface of data assimilation and machine learning](https://arxiv.org/abs/2610.07496) `2610.07496` | 0.50 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 32.6 | [Assumption-lean logistic regression with missing covariates](https://arxiv.org/abs/2610.07292) `2610.07292` | 0.50 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 32.1 | [Latent space bias directions in LLMs capture confidence, not fairness](https://arxiv.org/abs/2610.08559) `2610.08559` | 0.45 | 0.45 | 0.40 | 0.00 | 0.00 | — | — |
| 32.1 | [UNREAL: Unifying Retrieval and Long-Context with a Single Model](https://arxiv.org/abs/2610.08463) `2610.08463` | 0.30 | 0.40 | 0.30 | 0.00 | 0.50 | — | — |
| 32.1 | [SAGE: Semantic Anchor-Guided Evolution for Grounded Medical QA Data Synthesis](https://arxiv.org/abs/2610.08093) `2610.08093` | 0.45 | 0.05 | 0.30 | 1.00 | 0.00 | — | — |
| 32.1 | [HuatuoGPT-3: RL-Only Domain Adaptation from Base Models](https://arxiv.org/abs/2610.05966) `2610.05966` | 0.00 | 0.40 | 0.40 | 1.00 | 0.30 | — | — |
| 31.6 | [ConEx: Human-Interpretable Saliency Maps via Concept-Aware Attribution](https://arxiv.org/abs/2610.04605) `2610.04605` | 0.50 | 0.45 | 0.30 | 0.00 | 0.00 | — | — |
| 31.1 | [Disentangling Paradigm, Identifier, and Decoding in Generative Retrieval](https://arxiv.org/abs/2610.08716) `2610.08716` | 0.35 | 0.40 | 0.40 | 0.00 | 0.20 | — | — |
| 30.7 | [DiffGate: Difficulty-Gated Teacher Guidance for On-Policy Distillation](https://arxiv.org/abs/2610.04596) `2610.04596` | 0.30 | 0.40 | 0.40 | 0.00 | 0.28 | — | — |
| 30.5 | [Learning PDE solution operators with variable initial conditions via Latent Dynamics Networks](https://arxiv.org/abs/2610.08475) `2610.08475` | 0.50 | 0.40 | 0.30 | 0.00 | 0.00 | — | — |
| 30.5 | [Conversation Is a Two-Body Problem: Dyadic Evaluation of Full-Duplex Dialogue Models](https://arxiv.org/abs/2610.08125) `2610.08125` | 0.50 | 0.40 | 0.30 | 0.00 | 0.00 | — | — |
| 30.5 | [DirectSpeech2LLM: A Simple End-to-End Framework to Mitigate Prompt Overfitting in Speech-LLMs](https://arxiv.org/abs/2610.08085) `2610.08085` | 0.30 | 0.40 | 0.60 | 0.00 | 0.00 | — | — |
| 30.5 | [Self-Retrospection Distillation: Turning Post-hoc Experiences into Prior Foresight](https://arxiv.org/abs/2610.08077) `2610.08077` | 0.50 | 0.40 | 0.30 | 0.00 | 0.00 | — | — |
| 30.0 | [Toward Alignment Scaling Laws: A Framework and First Preregistered Measurements](https://arxiv.org/abs/2610.08540) `2610.08540` | 0.15 | 0.75 | 0.45 | 0.00 | 0.00 | — | — |
| 29.5 | [IdeaAnchor: Teaching LLMs to Turn Literature into Research Ideas](https://arxiv.org/abs/2610.08781) `2610.08781` | 0.30 | 0.40 | 0.40 | 0.00 | 0.20 | — | — |
| 29.5 | [Memory Depth and Reconstructed Context Width: A Controlled Evaluation of Hierarchical Retrieval](https://arxiv.org/abs/2610.08300) `2610.08300` | 0.30 | 0.40 | 0.40 | 0.00 | 0.20 | — | — |
| 29.5 | [Penalty-Framed No-Valid-Option MCQA: Analyzing LLM Abstention under Invalid Choices](https://arxiv.org/abs/2610.08153) `2610.08153` | 0.30 | 0.40 | 0.40 | 0.00 | 0.20 | — | — |
| 29.5 | [ShanLiangRen: A Nutrition Agent for Personalized Daily Meal Planning](https://arxiv.org/abs/2610.07886) `2610.07886` | 0.30 | 0.40 | 0.40 | 0.00 | 0.20 | — | — |
| 29.5 | [Trustworthy Method Comparison with AI Judges: Estimation and Design under Order, Batch, and Aggregation Effects](https://arxiv.org/abs/2610.07755) `2610.07755` | 0.30 | 0.40 | 0.40 | 0.00 | 0.20 | — | — |
| 28.9 | [A perspective note on likelihood approximation and inference for complex simulation models using a chain of aggregated normalizing flows](https://arxiv.org/abs/2610.07391) `2610.07391` | 0.65 | 0.40 | 0.00 | 0.00 | 0.00 | — | — |
| 28.9 | [TRACE: Rollout-Guided Quantization-Aware Training for FP4 Reinforcement Learning of MoE Language Models](https://arxiv.org/abs/2610.07767) `2610.07767` | 0.30 | 0.40 | 0.30 | 0.00 | 0.30 | — | — |
| 28.4 | [The Missing Minimal Pair: Stereotype Evaluation in LLMs](https://arxiv.org/abs/2610.08747) `2610.08747` | 0.30 | 0.40 | 0.00 | 1.00 | 0.00 | — | — |
| 28.4 | [Evidence-Bound Reasoning: Neuro-Semantic Verification of Biomedical AI in Glioblastoma Radiogenomics](https://arxiv.org/abs/2610.08660) `2610.08660` | 0.30 | 0.45 | 0.45 | 0.00 | 0.00 | — | — |
| 28.4 | [Harness Engineering for Software Engineering via Modular Executable Dev-Primitives](https://arxiv.org/abs/2610.07832) `2610.07832` | 0.30 | 0.45 | 0.30 | 0.00 | 0.20 | — | — |
| 27.4 | [Secure Speculative Decoding for Large Language Models](https://arxiv.org/abs/2610.08678) `2610.08678` | 0.30 | 0.45 | 0.40 | 0.00 | 0.00 | — | — |
| 27.4 | [DBRAG: Multi-Table Retrieval-Augmented Generation for Complex Database Queries](https://arxiv.org/abs/2610.07622) `2610.07622` | 0.30 | 0.40 | 0.30 | 0.00 | 0.20 | — | — |
| 26.8 | [GUI-HARVEST: Self-Improving GUI Agents through Evidence-Driven Harness Evolution](https://arxiv.org/abs/2610.00948) `2610.00948` | 0.00 | 0.00 | 0.40 | 1.00 | 0.50 | — | — |
| 26.3 | [SquidAgent: Parallelize Wisely, Coordinate Efficiently](https://arxiv.org/abs/2610.08647) `2610.08647` | 0.30 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 26.3 | [How Learning Governs Unlearning across the Memorization-Generalization Spectrum](https://arxiv.org/abs/2610.08577) `2610.08577` | 0.30 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 26.3 | [Agreement Is Not Validity: Cross-Model LLM Consensus in Diagnosing Student Failure Modes in K-12 Math Tutoring Dialogue](https://arxiv.org/abs/2610.08703) `2610.08703` | 0.30 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 26.3 | [Same-Number Citation Swaps: Stress-Testing Jev as a Financial Evidence Judge](https://arxiv.org/abs/2610.08675) `2610.08675` | 0.30 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 26.3 | [InterCorrect: Intersection-Aware Correction of Demographic Model Merging for Fair ASR](https://arxiv.org/abs/2610.08604) `2610.08604` | 0.30 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 26.3 | [Incidental information contaminates patient notes and disrupts clinical reasoning in large language models](https://arxiv.org/abs/2610.08585) `2610.08585` | 0.30 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 26.3 | [Wiki-Talkie: Multilingual Benchmarking of Persona-Based Agents on Real-World Discussions](https://arxiv.org/abs/2610.08513) `2610.08513` | 0.30 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 26.3 | [HINTT Submission to the 2nd MLC-SLM Challenge: Comparing Cascaded and Unified Approaches to Diarization and ASR](https://arxiv.org/abs/2610.08063) `2610.08063` | 0.30 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 26.3 | [Are Language Models Script-Aware?](https://arxiv.org/abs/2610.08037) `2610.08037` | 0.30 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 26.3 | [Adversarially Trained Linear Transformers Are Optimal Robust In-Context Learners for Gaussian Mixtures](https://arxiv.org/abs/2610.07754) `2610.07754` | 0.20 | 0.40 | 0.55 | 0.00 | 0.00 | — | — |
| 25.8 | [How Bregman Divergences Shape Shampoo](https://arxiv.org/abs/2610.08534) `2610.08534` | 0.25 | 0.40 | 0.30 | 0.00 | 0.20 | — | — |
| 25.8 | [Uniform Discrete Diffusion Models are Minimax Optimal for Estimating Distributions with Small Effective Support Size](https://arxiv.org/abs/2610.07655) `2610.07655` | 0.35 | 0.40 | 0.30 | 0.00 | 0.00 | — | — |
| 25.8 | [Does Muon Need Fine-Grained Spectral Shaping?](https://arxiv.org/abs/2610.07497) `2610.07497` | 0.25 | 0.40 | 0.30 | 0.00 | 0.20 | — | — |
| 25.3 | [Symphony for Text Generation: Benchmarking Clinical Note Generation](https://arxiv.org/abs/2610.08161) `2610.08161` | 0.30 | 0.45 | 0.30 | 0.00 | 0.00 | — | — |
| 25.3 | [From Delivery to Stateful Exploration: Rethinking the Index for Agentic Search](https://arxiv.org/abs/2610.07960) `2610.07960` | 0.30 | 0.45 | 0.30 | 0.00 | 0.00 | — | — |
| 25.3 | [CroissantMiner: Automated Extraction and Validation of Croissant Metadata for ML Datasets](https://arxiv.org/abs/2610.07132) `2610.07132` | 0.30 | 0.45 | 0.30 | 0.00 | 0.00 | — | — |
| 25.3 | [Personal-Agent Mediated Recommendation with Cross-Platform User History](https://arxiv.org/abs/2610.07588) `2610.07588` | 0.30 | 0.05 | 0.40 | 0.00 | 0.40 | — | — |
| 24.7 | [QF3: Fast Flow RL with Filtered Q-Gradients](https://arxiv.org/abs/2610.08789) `2610.08789` | 0.25 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 24.7 | [Adaptive Mean Estimation by In-Context Learning: A Gradient-Flow Analysis](https://arxiv.org/abs/2610.07804) `2610.07804` | 0.25 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 24.7 | [AGO AI Quality Gate: Evidence-First Release Decisions for Retrieval-Augmented Generation](https://arxiv.org/abs/2610.01218) `2610.01218` | 0.30 | 0.05 | 0.30 | 0.00 | 0.50 | — | — |
| 24.2 | [Conformal Prediction Sets Quantify Information Gain: A Theoretical Perspective](https://arxiv.org/abs/2610.08785) `2610.08785` | 0.50 | 0.40 | 0.00 | 0.00 | 0.00 | — | — |
| 24.2 | [POLAR: Ontology-Guided Risk Prevention for Tool-Calling LLM Agents](https://arxiv.org/abs/2610.08082) `2610.08082` | 0.30 | 0.40 | 0.30 | 0.00 | 0.00 | — | — |
| 24.2 | [Aligning Performance with Contribution: Towards Contribution-Aware Fair Recommendation](https://arxiv.org/abs/2610.08245) `2610.08245` | 0.50 | 0.40 | 0.00 | 0.00 | 0.00 | — | — |
| 24.2 | [Behavior-Mining, Generative Conversations, and Collaborative Advisory: the Future of Travel and Tourism Recommender Systems](https://arxiv.org/abs/2610.08232) `2610.08232` | 0.50 | 0.40 | 0.00 | 0.00 | 0.00 | — | — |
| 23.7 | [SlimWise: Decoupling Expert Pruning Across Prefill and Decode for Efficient MoE Serving](https://arxiv.org/abs/2609.34117) `2609.34117` | 0.00 | 0.45 | 0.45 | 0.00 | 0.30 | — | — |
| 23.2 | [HLA: Expressive Hybrid Linear Attention via Chunk-Wise Dynamic Mixing](https://arxiv.org/abs/2610.05842) `2610.05842` | 0.00 | 0.40 | 0.40 | 0.00 | 0.40 | — | — |
| 23.0 | [Towards In-Parameter Memory Augmentation for Large Language Models](https://arxiv.org/abs/2610.08630) `2610.08630` | 0.05 | 0.40 | 0.00 | 1.00 | 0.16 | — | — |
| 22.6 | [From Shared Demand Patterns to Local Uncertainty: Probabilistic Load Forecasting by Mixing Compact Adaptations](https://arxiv.org/abs/2610.08538) `2610.08538` | 0.25 | 0.40 | 0.30 | 0.00 | 0.00 | — | — |
| 21.6 | [Valid for Free: Homophily-Gated Conformal Prediction for Training-Free Node Classification with Tabular Foundation Models](https://arxiv.org/abs/2610.08564) `2610.08564` | 0.15 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 18.9 | [Language Unalignability: Why Some Concepts Resist Cross-Cultural Benchmark Evaluation](https://arxiv.org/abs/2610.08303) `2610.08303` | 0.30 | 0.45 | 0.00 | 0.00 | 0.00 | — | — |
| 18.4 | [Principled Under Pressure: Post-Training Decides Whether LLMs Act on Their Own Moral Judgment](https://arxiv.org/abs/2610.08670) `2610.08670` | 0.05 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |
| 17.9 | [Natural Language Questions as an Interface for Knowledge Graphs: QRAKEN Graph Distillation and Semantic Self-Healing](https://arxiv.org/abs/2610.08095) `2610.08095` | 0.00 | 0.45 | 0.40 | 0.00 | 0.00 | — | — |
| 16.8 | [Do Language Models Need a Trainable Input Embedding Table? Fixed Minimal Token Codes at 1.7B-Class Scale](https://arxiv.org/abs/2610.04002) `2610.04002` | 0.00 | 0.40 | 0.40 | 0.00 | 0.00 | — | — |

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