Multiple Gradient Descent Algorithm (MGDA)

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

• Gradient Descentmethod0.810
  Used for updating hidden state expectations; provides dynamical process theory testable against neuronal data
• Multi-Agent Deep Deterministic Policy Gradient (MADDPG)method0.758
  RL algorithm used to train baseline agents in the physical deception environment
• Gradient-based data attributionmethod0.750
  Baseline method against which probe-based ranking is compared; more computationally expensive.
• Gradient methodmethod0.750
  Optimization technique that computes weight changes by following the gradient of an error function; contrasted with evolutionary stochastic search.
• Gradient Descent on Free Energymethod0.743
  Optimization procedure for simultaneously updating action selection and perception; uses step size ζ (default 4).
• Gradient Descent Rotation Optimizationmethod0.732
  DAS uses SGD over differentiable parameterizations of orthogonal matrices (via PyTorch) to find optimal distributed alignments.
• Multiple-choice evaluation method for PM trainingmethod0.732
  Using language model log probabilities of answer choices (A)/(B) to produce preference labels.
• We hypothesize that degraded generalization on benchmarks like MMLU may reflect the computational demands of the tasks.hypothesis0.731
  Connecting the paper's task-difficulty findings to prior observations of weak generalization on complex QA benchmarks.