Labs Workspace

Active experiments
2
1 running · 1 queued
Evaluations run
38
This month
Datasets
4
Latest: reasoning-suite v4
Compute used
412GPU h
68% of monthly allocation

Model overview

Placeholder identifiers

Reference
apex-base v0.3
Baseline
Runs
14
Evals
12
Best
0.791
Evaluating
apex-base-ft v0.3.2
Domain-adapted
Runs
9
Evals
12
Best
0.847
Training
apex-staged v0.1
Staged reasoning
Runs
3
Evals
2
Best
0.812

Experiment tracking

Most recent runs

RunExperimentStatusProgressMetricDuration
EXP-0142
Staged reasoning, depth 3
apex-staged v0.1
Running
68%
0.8123 h 12 m
EXP-0141
Domain adaptation, low rank
apex-base-ft v0.3.2
Completed
100%
0.8475 h 40 m
EXP-0140
Retrieval grounding ablation
apex-base v0.3
Completed
100%
0.7912 h 05 m
EXP-0139
Long-context curriculum
apex-base-ft v0.3.1
Failed
41%
—1 h 18 m
EXP-0138
Verification pass, two-stage
apex-staged v0.1
Queued
0%
——
EXP-0137
Baseline re-run, seed 7
apex-base v0.3
Completed
100%
0.7741 h 52 m

EXP-0142 · training

Loss by step (thousands)

Live
3.02.21.50.70
02468
Training lossValidation loss

Evaluation results

Score by suite · higher is better

ReasoningRetrievalMultimodalCodeReliability
apex-base v0.3apex-base-ft v0.3.2apex-staged v0.1

Model comparison

apex-base v0.3apex-base-ft v0.3.2
  • Reasoning+13.0
  • Retrieval+8.0
  • Multimodal+5.0
  • Code+7.0
  • Reliability+3.0

Research projects

  • Staged reasoning62%
    AI reasoning
  • Evaluation that predicts use45%
    Model development
  • Document and chart understanding30%
    Multimodal intelligence
  • Routing for efficiency18%
    AI efficiency

Dataset overview

DatasetModalitySize
reasoning-suite v4
Updated 2 days ago
Text1,200 items
doc-understanding v2
Updated 6 days ago
Documents860 items
chart-qa v1
Updated 2 weeks ago
Image + text540 items
tool-use-traces v3
Updated 3 weeks ago
Structured3,400 traces

Development activity

  1. EXP-0142 reached step 6,800
    Validation metric 0.812 · 4 min ago
  2. Evaluation finished on apex-base-ft v0.3.2
    reasoning-suite v4 · 1,200 items · 38 min ago
  3. Training config updated
    Learning-rate schedule changed to cosine · 1 h ago
  4. Dataset reasoning-suite moved to v4
    86 items added, 12 corrected · 2 days ago
  5. EXP-0139 failed
    Out of memory at step 2,050 · 2 days ago

All runs, scores, datasets and model identifiers on this screen are illustrative sample data.