AI
Evidence-Grounded Auditing of Identification Assumptions in Climate-Policy Causal Evaluations
Publication: Hugging Face Daily Papers
Author: Not available in collected metadata
Source published: 24/09/2026, 20:00:00 UTC
Briefing created: 28/09/2026, 13:49:31 UTC
Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 28/09/2026, 11:18:24 UTC.
Briefings
ARGUS is a language-model pipeline designed to audit identification assumptions in climate-policy causal evaluations using an eleven-dimension rubric. Evaluated on 11-flaw benchmarks, it detects 73% of planted flaws versus 18% for keyword-based methods. Across 26 economics papers, it abstains on approximately 40% of assessments due to missing retrievable evidence. In a five-paper pilot, it assigned a higher risk level than the original labels on 25 of 33 completed assessments, though a pre-existing rule reduced this discrepancy. The tool generates evidence-linked risk reports to aid expert review without adjudicating causal claims.
Key points
- ARGUS detects 73% of planted flaws on the 11-flaw benchmark, significantly outperforming a keyword-based pipeline that detected only 18%.
- In a five-paper pilot, ARGUS assigned a higher risk level than the original labels on 25 of 33 assessments, though a pre-existing rule reduced this discrepancy.
Context and limitations
The pipeline abstains on approximately 40% of assessments across 26 economics papers due to a lack of retrievable evidence. In-sample discrepancies between assigned risk levels and original labels were reduced by a pre-existing rule, and causal claims remain unadjudicated.
Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 28/09/2026, 13:49:31 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.
View archived briefing on IPFS
bafybeicrlcbcgtfqcvveoxvbxkbxqoookri5twrifmknc744vis6opw6yi