Daily report — 2026-10-12
27 new papers screened · 10 analyzed · 0 reproduction runs · markdown
Reproductions
| status | paper | backend | claims matched |
|---|---|---|---|
| No reproduction runs today (opt-in). | |||
Top candidates
| score | paper | feasibility |
|---|---|---|
| 60.0 | Continually Post-training Small Models via Synthetic Self-distillation Continual post-training faces a dual challenge of acquisition and retention: models must learn new capabilities while preserving prior knowledge. While Self-Distillation Fine-Tuning (SDFT) mitigates this using self-supervision with in-context demonstrations, it relies on strong in-context learning and instruction-following abilities. Claimed: none extracted Evidence: + toy/synthetic experiments ("Synthetic"); + small/synthetic data ("Synthetic"); + public benchmarks ("benchmarks"); + explicit algorithm/pseudocode ("Algorithm 1"); no public code found; + keywords: sampling; + 39 HF upvotes | 7/10 heuristic |
| 56.8 | EmbodiedCLBench: Evaluating Continual Learning for Self-Evolving Embodied Agents Continual learning requires models to accumulate past experience and transfer it to new tasks. Despite substantial research on continual learning, existing work remains largely confined to non-embodied settings, without extending to embodied environments. Claimed: none extracted Evidence: + toy/synthetic experiments ("simulation"); + public benchmarks ("benchmark"); + claims simplicity ("simple"); + code: https://github.com/PKU-VaLuE-Lab/EmbodiedCLBench; + 62 HF upvotes | 6/10 heuristic |
| 55.3 | MetaKernelBench: Measuring GPU Kernel Knowledge Transfer Beyond Code Recent GPU kernel optimization agents retain what they learn in knowledge bases or as distilled skills. Kernel benchmarks score each attempt's implementation for correctness and speed but leave the reuse value of retained experience unmeasured. Claimed: none extracted Evidence: + public benchmarks ("benchmarks"); + formal statement ("bound"); + code: https://github.com/YIGE24/MetaKernelBench; + keywords: kernel; + 12 HF upvotes | 5/10 heuristic |
| 55.3 | S-Space: Exploring Spatial Workspace in Multimodal Language Models Multimodal language models (MLLMs) exhibit promising spatial reasoning capabilities, yet the internal mechanisms supporting them remain poorly understood. We uncover S-Space, a linear representation subspace in which object-token representations encode continuous coordinates along horizontal, vertical, and distance axes. Claimed: none extracted Evidence: + concrete method ("We propose a simple"); + claims simplicity ("simple"); + code: https://github.com/mirros-lab/s-space; + 197 HF upvotes | 6/10 heuristic |
| 54.2 | PreSTO: Predictive Subtree Prefetching for Fast LLM Power Sampling Power sampling improves large language model (LLM) reasoning at test time by sampling from a sharpened version of the base model's distribution. Its sampler, PowerMH, is slow because its Metropolis--Hastings (MH) updates are sequential: each accept/reject decision sets the state from which the next proposal is generated. Claimed: none extracted Evidence: - LLM-based ("LLM"); + formal statement ("We prove"); + code: https://github.com/Nan-Jiang-Group/PreSTO; + keywords: sampling, kernel, monte carlo; + 9 HF upvotes | 4/10 heuristic |
| 52.0 | The Life Cycle of a Massive Activation: Stochastic Birth, Weight-Decay-Driven Growth, and Competitive Consolidation Massive activations, residual-stream coordinates with magnitudes far larger than typical activations, are associated with attention sinks in transformers, but how their scale is regulated during training remains incompletely understood. Combining training-trajectory analyses and controlled interventions, we trace their emergence, growth, and consolidation. Claimed: none extracted Evidence: + formal statement ("bound"); + code: https://github.com/apd-smcclendon/massive_activations_2; + keywords: optimizer; + 2 HF upvotes | 5/10 heuristic |
| 51.1 | AMLworld-Compact: Importance-Weighted Downsampling for Cost-Effective LLM Evaluation and Error Diagnosis Evaluating large language models (LLMs) on AMLworld's smallest test splits is costly because each of over a million targets requires a separate graph prompt. Sampling reduces this cost, but small balanced subsets change unweighted precision and F1. Claimed: recall 1.0; F1 0.0 Evidence: - LLM-based ("LLM"); + explicit algorithm/pseudocode ("Algorithm 1"); + code: https://github.com/nat-nischw/AMLworld-Compact; + keywords: sampling; + 11 HF upvotes | 5/10 heuristic |
| 50.0 | RSI-Anything: Unlocking Recursive Self-Improvement with World Models Recursive self-improvement (RSI) aims to transform interaction experience into persistent improvements, yet external feedback is costly and world-model predictions may become unreliable as policies evolve. We introduce RSI-Anything, a domain-general framework built on JEPA-Anything that uses predictive world models to guide action and learning-data selection. Claimed: none extracted Evidence: + concrete method ("update rule"); - system/framework (many moving parts) ("framework"); + code: https://github.com/Gen-Verse/RSI-Anything; + 28 HF upvotes | 5/10 heuristic |
| 48.3 | Coding-Agent Benchmarks Should Match Their Users' Task Flows The evaluation of coding agents generally strives to be as realistic as possible. In our study, we collect 4,782 agent sessions of real software engineers in JetBrains IDEs, which we call Production Sessions. Claimed: none extracted Evidence: + public benchmarks ("Benchmarks"); + code: https://github.com/JetBrains-Research/SWE-TaskFlow-Generation-Public; + 5 HF upvotes | 4/10 heuristic |
| 47.9 | Rethinking Relative Policy Differences Transfer of On-Policy Distillation On-policy distillation (OPD) trains a target policy on its own rollouts using token-level supervision from a stronger teacher, but typically relies on the teacher being substantially more capable than the target. Recent work relaxes this requirement by transferring the relative policy difference between an expert and a reference policy rather than the expert's absolute distribution. Claimed: none extracted Evidence: + code: https://github.com/KnowledgeXLab/A2-OPD; + 10 HF upvotes | 4/10 heuristic |