DelTA: Discriminative Token Credit Assignment for Reinforcem

Reinforcement learning from verifiable rewards is enhanced through a discriminative token credit assignment method that improves reward-based training

By Sarah Chen 8 min read 2026-05-20

More this week

Development
π-Bench: Evaluating Proactive Personal Assistant A

Proactive assistance in personal agent systems requires identifying hi

2026-05-19 5 min
Creative AI
Full Attention Strikes Back: Transferring Full Att

RTPurbo leverages intrinsic sparsity in full-attention LLMs to achieve

2026-05-16 5 min
Research
Forecasting Scientific Progress with Artificial In

Current AI systems demonstrate limited capability in predicting scient

2026-05-21 5 min
Ethics
codebase-memory-mcp

A high-performance code intelligence MCP server that indexes codebases

Today 5 min
Development
OmniRoute

OmniRoute is a free and open-source AI gateway that connects to 237+ A

Today 5 min
Creative AI
Agent Reach - Give your AI agent eyes to see the e

Agent Reach is a capability layer that enables AI Agents to access int

Today 5 min
Research
SkillSpector

SkillSpector is a security scanner designed to detect vulnerabilities,

Today 5 min
Ethics
TREK - Self-hosted Collaborative Travel Planner

TREK is a self-hosted, real-time collaborative travel planner with map

Today 5 min
Development
Programmatic Tool Calling (PTC) for Performance Op

Programmatic Tool Calling (PTC) allows Claude to write code that calls

2025-11-24 5 min
Creative AI
Tool search with embeddings

Scale Claude applications to thousands of tools using semantic embeddi

2025-11-24 5 min
Research
Automatic context compaction

Long-running agentic tasks can often exceed context limits. Tool heavy

2025-11-24 5 min
Ethics
Reproduce Claude's agentic search benchmark scores

Build a Messages API harness that reproduces published DeepSearchQA an

2026-06-25 5 min
Development
Context engineering: memory, compaction, and tool

A cookbook guide comparing context engineering strategies for long-run

2026-03-20 5 min