SWE-DIVE evaluates coding agents in multi-turn sessions
A user simulator and conversation validator run inside Harbor to score whether the agent solves the task.
38
Product
SWE-bench
- Mentions · 7d0
- Mentions · 30d3
- All mentions10
- First seenSep 1, 2026
- Last mentioned9d ago
Timeline
10 articles · newest firstHSPR recovers software-agent benchmark scores from small task subsets
At a 10% task budget on four SWE-bench family banks, it had the lowest MAE on three and the lowest MaxErr on all four.
36
Before They Can Solve: Predicting Post-Training Coding-Agent Performance from Base Models
Three screens rank base models by likely coding-agent performance after expensive post-training.
54
TokenCast: Forecasting Token Consumption During LLM Agent Execution
TokenCast forecasts LLM agent token use from segment costs, cutting error 14.5% across 96 combinations.
46
Shockingly Simple Self-retrospection Improves Agentic Models Without RL
Retrospection-only fine-tuning lifts Qwen3.5-4B on SWE-bench without rewards or an external teacher.
58
Trajectory-Level Security Debt in LLM Coding Agents
A new metric tracks static-analysis security debt across intermediate states of LLM coding agents, not just final code.
46
AdaRepair-Mem: Adaptive Experience Orchestration for Repository-Level Program Repair
AdaRepair-Mem introduces coverage-, quality-, and stage-aware memory retrieval that improves LLM-based repository-level program repair on SWE-Bench.
28
Monitoring and Discovering Reward Hacking with Internal Representations during LLM Evaluations
Researchers use difference-of-means representation vectors to detect reward hacking in frontier LLMs; GLM 5.2 hacks 73% of SWE-bench rollouts.
55
What Does an LLM-Agent Leaderboard Rank Actually Compare?
A methodological study shows close LLM-agent leaderboard rank gaps on SWE-bench and similar benchmarks often do not support superiority claims.
20
Bilevel Coordinated Reflection: A Game-Theoretic Approach to Multi-Agent LLM Systems
Paper models multi-agent LLM orchestration as a bilevel game, proving transcript-only gating limits and introducing grounded-memory SRMA.
32
Appears with
- Qwen3.5-4Bai model · 1
- DataAgentBenchproduct · 1
- mini-SWE-agentproduct · 1
- Harborproduct · 1
- AgentRewardBenchproduct · 1
- GLM 5.2ai model · 1
- Qwen 3.8 Maxai model · 1
- DeepSWEproduct · 1
- MirrorCodeproduct · 1
- tau2-benchproduct · 1
- AdaRepair-Memproduct · 1
- SWE-DIVEproduct · 1
- Kimiai model · 1
- Open Agentproduct · 1
- Kimi K3ai model · 1
- ProgramBenchproduct · 1
Entities are extracted by the model from each article. First seen , last mentioned . Watching keeps it in this browser only (no account); the watchlist page and dashboard alerts use it.