Tuesday 15 September 2026Paper No. 1,386

Evening EditionRSS

ZOTPAPER

News without the noise

Agent Harnesses

Harnesses, orchestration, context engineering

61 articles

All Agent Harnesses Tools & MCP Multi-Agent Systems Creative Agents Memory & Planning Safety & Evals RL for Agents AI & ML

Viewgrid list compact

Compressing Shared Sandbox Memory Makes High-Fanout Agents More Efficient

Embedded software agents need interaction contracts and continuous assurance

VikingRAG Cuts Token Use for Retrieval Over Structured Documents

Similarity-aware context windows improve model routing across multi-turn conversations

Dual-layer knowledge graph connects fragmented pharmaceutical process-development documents

Cross-task failure diagnosis makes LLM agent harness training faster

Bandit-guided skill evolution cuts LLM agent optimization costs

Embedded AI agents can adapt software during live use

A semantic harness lets vision-language models control different robots

Scene-grounded decoding keeps vision-language plans executable and visually supported

Benchmark tests retrieval for agent-written queries across 190 million web pages

Adam-inspired updates make agent skill evolution steadier and more efficient

Controlled harness reveals what helps agents refactor entire repositories

TROVE edits agent workflows only when runtime evidence demands it

Training agents to combine graphical interfaces and commands improves efficiency

Personalizing agents through cross-session interaction data boosts task success

Persistent evidence tracking improves language-agent performance on bioinformatics benchmarks

Evolving terminal environments keeps training tasks challenging as agents improve

Distilled repository skills improve agents conducting machine-learning research

Civilization benchmark exposes agents’ weak monitoring and long-term follow-through

Branching Across Hypotheses Helps Research Agents Avoid Early Search Traps

Reusing agent trajectories cuts token use on repeated tool tasks

LLMs can build agent harnesses, but improvements transfer poorly

Typed trace folding helps agents and observers manage long runs

Layered agent harnesses improve software over repeated autonomous development cycles

Long-video agent adapts retrieval to each question’s evidence needs

Models struggle to improve themselves from vague capability goals

Models remain weak at steering coding agents through long tasks

Safe agent self-modification requires verifiable, expressive recovery mechanisms

Simulated user personas improve agents’ handling of multi-turn workflows

Tool-call traces expose extraction failures that source fidelity misses

Agents learn to manage long contexts through fine-grained reinforcement learning

Markdown interfaces cut agent context costs without reducing task success

DSA orchestrates evidence-aware stock research across six regional markets

Agent harnesses can elevate malicious context past instruction defenses

Behavior-aware checks make agent harness evolution cheaper and more reliable

Generating Agent Harnesses on Demand Improves Models Across Benchmarks

Self-evolving harness turns general AI agents into stronger RCA specialists

Stepwise LLM routing cuts multi-agent workflow costs while preserving quality

Full-context drafters help compressed agent models retain accuracy

Self-improving context programs help frozen models understand long videos

Literate programming links prose, code and symbols for coding agents

Separating state tracking from actions improves multi-turn tool reliability

Parallel trials and subtasks cut latency in difficult LLM reasoning

Persistent corpus maps help agents find evidence within tight budgets

Good tool plans can still fail under resource constraints

Coding agents extend LLM serving simulators with greater fidelity and speed

Bounded citation-graph search improves recall while cutting scholarly search costs

Switching Models Mid-Task Imposes a Cost and Quality Penalty

Evolved agent harnesses lift enterprise performance without retraining models

Selecting complementary LLM skills improves success while reducing context use

Adaptive task selection cuts the cost of optimizing LLM agent harnesses

Cost-aware model routing matches exhaustive estimation with fewer expensive checks

Python objects become AI agents in new NVIDIA framework

Re-evaluation shows harness evolution for agents may not outperform simple test-time scaling

Letting language models decide when to compact their context windows

Agent framework auto-tunes video diffusion for 2x speedup

Orchestrator models dynamically combine specialized LLMs into collective intelligence

Pruning context to recent tool calls improves agent reliability and efficiency

Self-supervised method improves agent harnesses using past trajectories

Proposing a workflow store to harden AI agents against failure