Microsoft and UC Berkeley release TokenEconomizer for coding agents
On 277 SWE-Bench tasks it cut mean cost 18.4% to 30.9% across five agent-model settings without changing weights.
46
Organization
Microsoft Research
- Mentions · 7d2
- Mentions · 30d3
- All mentions5
- First seenSep 2, 2026
- Last mentioned21h ago
Timeline
5 articles · newest firstMickens argues AI safety and hardware isolation at LABScon 26
He cites sandbox escapes and describes Guillotine, a hypervisor that isolates model CPU cores from hypervisor state.
36
Clearing chat history helps AI agents find better solutions
Microsoft Research shows long-running AI agents plateau in "idea basins"; a Fork-and-Flush context-reset method boosts scores 66% on equal compute.
48
ExecCritic: Learn to Test, Test to Improve for Coding Agents
ExecCritic separates test generation from patching for coding agents, lifting SWE-bench Verified resolution to 72.6%.
36
microsoft/VibeVoice-ASR-Streaming-7B — new model trending #27 on Hugging Face
Microsoft released VibeVoice-ASR-Streaming-7B, an open streaming ASR model with speaker attribution, custom hotwords, and support for 10 languages under MIT license.
40
Appears with
- Microsoftvendor · 2
- VibeVoice-ASR-Streaming-7Bai model · 1
- GPT-5.6 Solai model · 1
- TokenEconomizerproduct · 1
- GPT-5.6 Terraai model · 1
- GPT-5.6-solai model · 1
- UC Berkeleyorganization · 1
- Codexproduct · 1
- Harvard Universityorganization · 1
- Piproduct · 1
- Guillotineproduct · 1
- Qwen-3.5-35B-A3Bai model · 1
- Opus 4.8ai model · 1
- Claude Codeproduct · 1
- Sonnet 5ai model · 1
- SentinelLABSorganization · 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.