OctoLong: Mid-Training On Cross-Repository Code Contexts Enhances Long-Context Modeling
Paper • 2608.05141 • Published • 1
Dipankar Sarkar PRO
dipankarsarkar
AI & ML interests
Building the AI-native stack. Agents as infrastructure, safety as architecture, performance as plumbing. I publish the receipts: papers, datasets, demos.
Recent Activity
upvoted a paper about 2 hours ago
OctoLong: Mid-Training On Cross-Repository Code Contexts Enhances Long-Context Modeling published an article about 2 hours ago
quantal-ternary: a 0.5B model that learned honesty — from 11.34 to 0.5597 upvoted a paper about 3 hours ago
ADIAS: Automated Design of Interactive Agentic SystemsOrganizations
upvoted a paper about 2 hours ago
upvoted 6
papers about 3 hours ago
ADIAS: Automated Design of Interactive Agentic Systems
Paper • 2608.06410 • Published • 2
Scrouting: Cost-Aware Routing of Coding Agents by Scouting the Repository First
Paper • 2608.04804 • Published • 1
Attend to Your Own Thoughts: Breaking the Barrier for Post-Training Quantization of Reasoning LLMs through the Lens of 1.58-Bit Quantization
Paper • 2608.01078 • Published • 1
Spend Bits Where Queries Look: KV Cache Vector Quantization with Attention-Preserving Transforms
Paper • 2608.04074 • Published • 2
Harness-R1: Learning to Edit Executable Runtime Harnesses from Agent Failure Trajectories
Paper • 2608.02276 • Published • 4
Permission Denied: Policy-Graded Evaluation of Coding Agents in Hardened Environments
Paper • 2608.02670 • Published • 2
upvoted 4
papers about 4 hours ago
InsightEmb: Learning Action-Intent Embeddings for Agentic Insight Retrieval
Paper • 2608.04761 • Published • 1
NOVA: A Verification-Aware Agent Harness for Architecture Evolution in Industrial Recommender Systems
Paper • 2606.27243 • Published • 1
Living-Harness Is an Interactive-Agent Evolver
Paper • 2607.26598 • Published • 1
Antares: Foundation Models for Agentic Vulnerability Localization
Paper • 2608.02407 • Published • 2
upvoted 2
papers about 5 hours ago
upvoted 6
papers about 7 hours ago
DistilVDR: A Compact End-to-End Visual Document Retriever via Dual-Student Distillation
Paper • 2608.10636 • Published • 5
UniMoMo: Expert Merging-Based MoE Acceleration for Large Recommendation Models
Paper • 2608.08627 • Published • 6
OpenSkillRisk: Benchmarking Agent Safety When Using Real-World Risky Third-Party Skills
Paper • 2607.20121 • Published • 1
Isolation as a First-Class Principle for LLM-Agent System Safety: Concepts, Taxonomy, Challenges and Future Directions
Paper • 2607.12406 • Published • 1
Reasoning effort, not tool access, buys first-try reliability in agentic code generation: an observational study
Paper • 2607.02436 • Published • 1
Screenshots or Tools? Eliciting Tool Use and Managing Multimodal Context in Hybrid GUI-MCP Computer-Use Agents
Paper • 2608.03327 • Published • 3
upvoted a paper about 8 hours ago