diff --git a/convlab2/policy/README.md b/convlab2/policy/README.md index cc0b29edaaa66dde5febe62182ee709e800dab67..1990cdd6b03a38fe6f5a4f8b4eb8a9708761c590 100755 --- a/convlab2/policy/README.md +++ b/convlab2/policy/README.md @@ -16,3 +16,21 @@ The interfaces for dialog policy are defined in policy.Policy: and outputs the next system action. - **init_session** reset the model variables for a new dialog session. + +## Rule based simulator results + +| Model | Complete rate | Success rate | Average return | Turns | Average actions | +|-------|---------------|--------------|----------------|-------|-----------------| +| MLE | | | | | | +| PG | | | | | | +| GDPL | | | | | | +| PPO | | | | | | + +## Transformer based user simulator (TUS) results + +| Model | Complete rate | Success rate | Average return | Turns | Average actions | +|-------|---------------|--------------|----------------|-------|-----------------| +| MLE | | | | | | +| PG | | | | | | +| GDPL | | | | | | +| PPO | | | | | |