RUMP SCIENCE

Edison Scientific announces agreement with Pfizer

ai4sci-labs · OCT 29

REPORT: Korea's NanoForge AI raises $10.9M seed to automate materials discovery

gnews-discovery · OCT 29

AI framework mines traditional Chinese medicine for drug discovery

gnews-discovery · OCT 29

[PAPER] STUDY: Hidden dislocation hierarchies change how grain boundaries slide... GBClaw agentic AI framework used

arxiv-science · OCT 20

[PAPER] STUDY: Claude Opus tops published science models by 10.4%, but agents fall short of real insight...

arxiv-science · OCT 20

[PAPER] STUDY: Only 1 of 12 AI models discovers altered physics when left to explore freely... the other 11 score zero

arxiv-science · OCT 27

[PAPER] STUDY: MatOS agent beats Codex CLI and Claude Code on science benchmarks with the same base model

arxiv-science · OCT 25

[PAPER] New ClaimNov tool catches 97% of disguised papers at 1.5% false-alarm rate

arxiv-science · OCT 27

Anthropic expands AI for Science program to 20 research partners

gnews-autoresearch · OCT 29

Johns Hopkins cancer cell mapping tech picked for Anthropic AI for Science project

gnews-autoresearch · OCT 29

[PAPER] STUDY: Tabular foundation models match or beat bespoke galaxy-redshift estimators with just 1,000 examples

arxiv-science · OCT 29

[PAPER] STUDY: AI system finds compact 'optimization laws' behind human risky choice and central-bank policy

arxiv-science · OCT 20

[PAPER] Study: AI predicts electron densities so well that a single step replaces the full calculation

arxiv-science · OCT 29

[PAPER] STUDY: Qwen2.5-7B's early gain over an LLM-free loop did not persist in 50-seed replication

arxiv-science · OCT 17

[PAPER] STUDY: One interference type fooled the SETI detector 62.7% of the time

arxiv-science · OCT 15

[PAPER] STUDY: Poison the protein data, and the AI designs what the attacker wants... while behaving normally

arxiv-science · OCT 15

[PAPER] STUDY: AI scientists make partial progress, rarely finish: rubric scores top out at 59%

arxiv-science · OCT 29

[PAPER] STUDY: AI agents pick research questions, derive theory and test it on quantum spins in diamond

arxiv-science · OCT 29

[PAPER] STUDY: AI agents' forecasts of experiment outcomes no more accurate than a fresh copy of the same model

arxiv-science · OCT 29

[PAPER] STUDY: Smarter models aren't enough for AI science... the orchestration layer matters too

arxiv-science · OCT 13

[PAPER] Paper: LLM agent designs algorithms and proves their runtime bounds... no empirical comparison in the pipeline

arxiv-science · OCT 26

[PAPER] STUDY: Agent supervising spectrum analysis recovers 16 of 20 failures, returns NaN on 4 it won't trust...

arxiv-science · OCT 19

[PAPER] STUDY: Autonomous research agents tested on 15 medical tasks... competitive scores, shaky risk estimates

arxiv-science · OCT 18

[PAPER] Self-building labs: AI that designs, builds and runs its own laboratory from an empty room

arxiv-science · OCT 16

[PAPER] STUDY: Agent harness can increase harmful compliance in GPT models, even under direct prompting...

arxiv-science · OCT 23

[PAPER] STUDY: AI-assisted judge cuts false-positive math proof verdicts from 23% to 6%

arxiv-science · OCT 29

[PAPER] STUDY: Automated detector flags research-integrity violations in 99.7% of 300 AI research-agent trajectories

arxiv-science · OCT 22

[PAPER] STUDY: Three apparent AI model failures shrink or vanish once benchmark controls are applied

arxiv-science · OCT 22

[PAPER] New system gives ranked candidate years for 56 undated entries of the Babylonian Astronomical Diaries...

arxiv-science · OCT 25

[PAPER] STUDY: Neural-network wavefunctions find hints of exotic parton phases at ν=4/3 and 7/5...

arxiv-science · OCT 28

[PAPER] STUDY: ReactNexus AI maps chemical reaction networks from reactant graphs alone, no templates needed

arxiv-science · OCT 28