CHANNEL ARCHIVE // [345] ARTICLES // PAGE 19 OF 23
Artificial Intelligence
Research into artificial intelligence and intelligent systems.
CHANNEL // 01 [345]
Artificial Intelligence
Research into artificial intelligence and intelligent systems.
When AI Conversations Go Off Track, Rebuilding Context Can Help
Source published
- large language models
- multi-turn conversations
- context pollution
When Can Fast Decision Models Replace LLM Judges?
Source published
- LLM-as-a-judge
- System-1 decision models
- Jev
TAILOR targets the risky tail of AI-generated computer-system heuristics
Source published
- large language models
- heuristic synthesis
- TAILOR
Swapping Answer Options Exposes Hidden Weaknesses in AI Decision Models
Source published
- System One decision models
- demographic bias
- AI auditing
Estimating AI Reliability From Fewer Repeated Trials
Source published
- AI evaluation
- Beta mixtures
- repeated sampling
Compressing Small Language Models: When Combining Techniques Helps—and When It Fails
Source published
- small language models
- model compression
- quantization
Exo lets AI agents rewrite their harness while working
Source published
- Exo
- AI agents
- agent harnesses
Medical AI scores fall sharply when answer choices disappear
Source published
- medical AI
- large language models
- medical question answering
Matching Score Distributions Could Help AI Judges Align With Human Ratings
Source published
- LLM-as-a-judge
- streaming quantile matching
- score calibration
A framework for human–AI teamwork that accounts for how advice changes decisions
Source published
- human–AI collaboration
- collaborative omnimechanism design
- outcome indistinguishability
A Transformer’s Arithmetic Circuit Works Despite Local Geometric Distortion
Source published
- mechanistic interpretability
- modular addition
- transformer circuits
When AI Gets the Score Right but the Work Wrong
Source published
- AI monitoring
- reward hacking
- large language models
Behavior graphs turn robot deployment recordings into maps of success and failure
Source published
- robotics
- behavior graphs
- deployment data
CalVLA makes robot confidence more informative—but not control more successful
Source published
- CalVLA
- vision-language-action policies
- OpenVLA
AI Safeguards Need to Be Judged by How They Fail Together
Source published
- AI agents
- LLM safeguards
- correlated errors