SourceJul 3, 20262026-07-034 minVerified
Combining RLVR and Human Demonstrations for Better LMs
A paper on combining RLVR with human demonstrations to train style, structure, and diversity beyond verifiable rewards.
CommunityJul 3, 20262026-07-035 minVerified
How To Compare Code Models Beyond Benchmark Scores
Code model evaluation should weigh real task success, retries, latency, and token cost, not benchmark scores alone.
SourceJul 3, 20262026-07-034 minVerified
Conditional Co-Ablation Reveals Hidden Backup Transformer Circuits
How CoAx exposes backup circuits that single ablation can miss due to self-repair in transformers.
SourceJul 3, 20262026-07-036 minVerified
Context Governance for Verifiable AI Agent Knowledge
How ContextNest frames context governance with a verifiable knowledge vault layer for auditable AI agents beyond retrieval quality.
SourceJul 3, 20262026-07-034 minVerified
DiscoLoop Tests Internal Multi-Hop Reasoning In One Pass
DiscoLoop explores multi-hop reasoning inside a single forward pass without relying on long external CoT tokens.
SourceJul 3, 20262026-07-034 minVerified
Can Multimodal AI Improve Rail Crossing Safety Assessment
Examines whether combining rail crossing images with accident records improves safety assessment and what validation matters.
SourceJul 3, 20262026-07-034 minVerified
OCB Tests Native Office Understanding Beyond PDF QA
OCB evaluates native Office file understanding, revealing document AI limits beyond PDF-based QA.
CommunityJul 3, 20262026-07-034 minVerified
Why Foundational Learning Still Matters in the AI Era
AI can boost productivity but also amplify errors, making foundational learning essential for problem framing, verification, and judgment.
RoundupJul 2, 20262026-07-022 min
AI Resource Roundup (24h) - 2026-07-02
A curated link roundup from recently collected official updates and tech news.
SourceJul 2, 20262026-07-024 minVerified
Bug Reproduction Tests as Signals for Code Agents
How code agents can use bug reproduction tests as diagnostic signals during patch generation, not just post-hoc checks.
SourceJul 2, 20262026-07-023 minVerified
DART-VLN Improves Discrete VLN Without Retraining at Test Time
DART-VLN targets stale memory reads and local backtracking in discrete VLN using training-free test-time control.
SourceJul 2, 20262026-07-025 minVerified
Dynamic 3D Reconstruction from Monocular Video with Generative Priors
A method for building dynamic 3D Gaussians from monocular video and correcting reconstruction gaps with a conditional video model.
SourceJul 2, 20262026-07-024 minVerified
Interpreting RAG Retrieval With Sparse Autoencoder Features
Explores using sparse autoencoders to disentangle dense RAG embeddings for interpretable retrieval analysis and steering.
SourceJul 2, 20262026-07-024 minVerified
Latent Space Control for Trustworthy LLM Behavior
From steering vectors to model calibrators, this paper frames latent-space intervention as a path to better LLM control and trust.
CommunityJul 2, 20262026-07-024 minVerified
Which Jobs Are Safer: Office or Skilled Trades?
Official data on AI and automation exposure compares office jobs and skilled trades by task structure and employment outlook.
SourceJul 2, 20262026-07-023 minVerified
On-Device AI Security Across App, Model, and OS
A look at the main security risks in mobile on-device AI, focusing on attack surfaces across apps, models, and OS.
CommunityJul 2, 20262026-07-025 minVerified
Public AI Infrastructure: Distributed Access or Concentrated Scale
Examines distributed vs. concentrated public AI compute strategies and what they mean for sovereign AI capacity.
CommunityJul 2, 20262026-07-024 minVerified
How Safety Holds Up Across Long AI Conversations
Examines whether AI safety remains consistent in long conversations and highlights gaps in session-level evaluation.
SourceJul 2, 20262026-07-023 minVerified
Scaling Thermodynamic AI With Backprop and Gibbs Sampling
How Ising-based thermodynamic computing may scale training, with focus on sampling costs and hardware limits.
SourceJul 2, 20262026-07-024 minVerified
Staleness and Learning Rates in Asynchronous RLHF
Examines how stale rollouts and learning rates affect stability in asynchronous RLHF, with practical signals like staleness and ESS.
CommunityJul 1, 20262026-07-015 minVerified
AI Employment Narrative Shifts From Loss to Redesign
Examines whether AI eliminates jobs or redesigns tasks, and why this shift matters for hiring, reskilling, and productivity.
RoundupJul 1, 20262026-07-012 min
AI Resource Roundup (24h) - 2026-07-01
A curated link roundup from recently collected official updates and tech news.
SourceJul 1, 20262026-07-015 minVerified
Design Axes for Agentic Orchestration in Enterprises
A practical guide to balancing agent autonomy, traceability, and control in enterprise orchestration design.
SourceJul 1, 20262026-07-014 minVerified
Emergent Misalignment Depends on More Than Training Data
EM may depend on optimizers and batch settings, making finetuning recipes part of safety evaluation, not just data.