AI Paper Atlas

Research map / Probabilistic and scientific modeling / Causal discovery

Counterfactual Maximum Likelihood Estimation for Training Deep Networks

Xinyi Wang, Wenhu Chen, Michael Saxon, William Yang Wang

NeurIPS 2021 · Probabilistic and scientific modeling › Causal discovery · approach: Latent-variable causal estimation · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
train with Implicit or Explicit CMLE instead of MLEMLP · simulated data OOD2MSE: 7.25 -> 6.94Improved
Explicit CMLE with p(T|X,Y) trained on interventional distribution (Explicit CMLE*) (ablation)Explicit CMLE · simulated counterfactual testMSE: 4.77 -> 4.61Improved
Explicit instead of Implicit CMLE (ablation)CMLE · simulated counterfactual testMSE: 5.11 -> 4.77Improved
CMLE vs MLE on in-distribution observational testCMLE · simulated observational testMSE: 3.63 -> 3.45Improved
fine-tune RoBERTa on counterfactual-augmented data with Implicit CMLERoBERTa · ANLIaccuracy: 31.3 -> 32.2Improved
augment with counterfactual data using plain MLE generator (ablation)RoBERTa · ANLIaccuracy: 31.3 -> 30.8Hurt
CMLE augmentation on in-domain NLIRoBERTa · SNLIaccuracy: 92.6 -> 92.4No clear effect
CMLE generators vs MLE for hypothesis label consistencyBART · SNLI generated hypotheses (NLI-RoBERTa)label consistency: 65.5 -> 95.2Improved
Explicit instead of Implicit CMLE for generation consistency (ablation)BART · MNLI generated hypotheses (NLI-RoBERTa)label consistency: 75.1 -> 85.8Improved
train captioner on hallucination-augmented data with plain MLE (ablation)Transformer captioner · MSCOCO Karpathy testCIDEr: 106.7 -> 102.1Hurt
Implicit CMLE vs MLE on hallucinationTransformer captioner · MSCOCO Karpathy testCHAIRs: 7.3 -> 6.8Improved
Implicit CMLE vs MLE on similarity metricsTransformer captioner · MSCOCO Karpathy testCIDEr: 106.7 -> 105.2No clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy