AMAGO: Scalable In-Context Reinforcement Learning for Adaptive Agents
AMAGO: Scalable In-Context Reinforcement Learning
for Adaptive Agents
Jake Grigsby1
Jim Fan2
@@ -193,7 +193,7 @@
@@ -217,7 +217,7 @@
-
+
@@ -232,7 +232,7 @@
-
+
@@ -259,7 +259,7 @@
-
+
@@ -318,7 +318,7 @@
In-context RL is applicable to any memory, generalization, or meta-learning problem, and we have designed AMAGO to be flexible enough to support all of these cases. Our code is fully open-source and includes examples of how to apply AMAGO to new domains. We hope our agent can serve as a strong baseline in the development of new benchmarks that require long-term memory and adaptation. Check it out on GitHub here.
-
+
diff --git a/_site/src/figure/case_studies_arxiv_v2.pdf b/_site/src/figure/case_studies_arxiv_v2.pdf
deleted file mode 100644
index 291c3a6..0000000
Binary files a/_site/src/figure/case_studies_arxiv_v2.pdf and /dev/null differ
diff --git a/_site/src/figure/case_studies_arxiv_v2.png b/_site/src/figure/case_studies_arxiv_v2.png
new file mode 100644
index 0000000..98893e5
Binary files /dev/null and b/_site/src/figure/case_studies_arxiv_v2.png differ
diff --git a/_site/src/figure/crafter_condensed_results.pdf b/_site/src/figure/crafter_condensed_results.pdf
deleted file mode 100644
index 7177512..0000000
Binary files a/_site/src/figure/crafter_condensed_results.pdf and /dev/null differ
diff --git a/_site/src/figure/crafter_condensed_results.png b/_site/src/figure/crafter_condensed_results.png
new file mode 100644
index 0000000..c17e445
Binary files /dev/null and b/_site/src/figure/crafter_condensed_results.png differ
diff --git a/_site/src/figure/fig1_iclr_e_notation.pdf b/_site/src/figure/fig1_iclr_e_notation.pdf
deleted file mode 100644
index 63fc3e8..0000000
Binary files a/_site/src/figure/fig1_iclr_e_notation.pdf and /dev/null differ
diff --git a/_site/src/figure/fig1_iclr_e_notation.png b/_site/src/figure/fig1_iclr_e_notation.png
new file mode 100644
index 0000000..86f80e9
Binary files /dev/null and b/_site/src/figure/fig1_iclr_e_notation.png differ
diff --git a/_site/src/figure/popgym_summary_expanded_outliers.pdf b/_site/src/figure/popgym_summary_expanded_outliers.pdf
deleted file mode 100644
index 909991a..0000000
Binary files a/_site/src/figure/popgym_summary_expanded_outliers.pdf and /dev/null differ
diff --git a/_site/src/figure/popgym_summary_expanded_outliers.png b/_site/src/figure/popgym_summary_expanded_outliers.png
new file mode 100644
index 0000000..7d8d5db
Binary files /dev/null and b/_site/src/figure/popgym_summary_expanded_outliers.png differ
diff --git a/_site/src/logos/rpl_logo.pdf b/_site/src/logos/rpl_logo.pdf
deleted file mode 100644
index ce3826d..0000000
Binary files a/_site/src/logos/rpl_logo.pdf and /dev/null differ
diff --git a/_site/src/logos/rpl_logo.png b/_site/src/logos/rpl_logo.png
new file mode 100644
index 0000000..89195ff
Binary files /dev/null and b/_site/src/logos/rpl_logo.png differ
diff --git a/index.markdown b/index.markdown
index 17b685e..cce3a08 100644
--- a/index.markdown
+++ b/index.markdown
@@ -158,7 +158,7 @@ highlight {
-AMAGO: Scalable In-Context Reinforcement Learning for Adaptive Agents
+AMAGO: Scalable In-Context Reinforcement Learning
for Adaptive Agents
Jake Grigsby1
Jim Fan2
@@ -204,7 +204,7 @@ AMAGO improves memory and adaptation by optimizing long-context Transformers on
@@ -228,7 +228,7 @@ In-Context RL's flexibility lets us evaluate AMAGO on many generalization, memor
-
+
@@ -243,7 +243,7 @@ AMAGO handles meta-learning as a simple extension of zero-shot generalization, a
-
+
@@ -270,7 +270,7 @@ As an example, we evaluate instruction-conditioned agents in the procedurally ge
-
+
@@ -329,7 +329,7 @@ Above, we use several single-task instructions to evaluate the exploration capab
In-context RL is applicable to any memory, generalization, or meta-learning problem, and we have designed AMAGO to be flexible enough to support all of these cases. Our code is fully open-source and includes examples of how to apply AMAGO to new domains. We hope our agent can serve as a strong baseline in the development of new benchmarks that require long-term memory and adaptation. Check it out on GitHub here.
-
+
diff --git a/src/figure/case_studies_arxiv_v2.pdf b/src/figure/case_studies_arxiv_v2.pdf
deleted file mode 100644
index 291c3a6..0000000
Binary files a/src/figure/case_studies_arxiv_v2.pdf and /dev/null differ
diff --git a/src/figure/case_studies_arxiv_v2.png b/src/figure/case_studies_arxiv_v2.png
new file mode 100644
index 0000000..98893e5
Binary files /dev/null and b/src/figure/case_studies_arxiv_v2.png differ
diff --git a/src/figure/crafter_condensed_results.pdf b/src/figure/crafter_condensed_results.pdf
deleted file mode 100644
index 7177512..0000000
Binary files a/src/figure/crafter_condensed_results.pdf and /dev/null differ
diff --git a/src/figure/crafter_condensed_results.png b/src/figure/crafter_condensed_results.png
new file mode 100644
index 0000000..c17e445
Binary files /dev/null and b/src/figure/crafter_condensed_results.png differ
diff --git a/src/figure/fig1_iclr_e_notation.pdf b/src/figure/fig1_iclr_e_notation.pdf
deleted file mode 100644
index 63fc3e8..0000000
Binary files a/src/figure/fig1_iclr_e_notation.pdf and /dev/null differ
diff --git a/src/figure/fig1_iclr_e_notation.png b/src/figure/fig1_iclr_e_notation.png
new file mode 100644
index 0000000..86f80e9
Binary files /dev/null and b/src/figure/fig1_iclr_e_notation.png differ
diff --git a/src/figure/popgym_summary_expanded_outliers.pdf b/src/figure/popgym_summary_expanded_outliers.pdf
deleted file mode 100644
index 909991a..0000000
Binary files a/src/figure/popgym_summary_expanded_outliers.pdf and /dev/null differ
diff --git a/src/figure/popgym_summary_expanded_outliers.png b/src/figure/popgym_summary_expanded_outliers.png
new file mode 100644
index 0000000..7d8d5db
Binary files /dev/null and b/src/figure/popgym_summary_expanded_outliers.png differ
diff --git a/src/logos/rpl_logo.pdf b/src/logos/rpl_logo.pdf
deleted file mode 100644
index ce3826d..0000000
Binary files a/src/logos/rpl_logo.pdf and /dev/null differ
diff --git a/src/logos/rpl_logo.png b/src/logos/rpl_logo.png
new file mode 100644
index 0000000..89195ff
Binary files /dev/null and b/src/logos/rpl_logo.png differ
-
+
@@ -259,7 +259,7 @@
-
+
@@ -318,7 +318,7 @@
In-context RL is applicable to any memory, generalization, or meta-learning problem, and we have designed AMAGO to be flexible enough to support all of these cases. Our code is fully open-source and includes examples of how to apply AMAGO to new domains. We hope our agent can serve as a strong baseline in the development of new benchmarks that require long-term memory and adaptation. Check it out on GitHub here.
-
+
diff --git a/_site/src/figure/case_studies_arxiv_v2.pdf b/_site/src/figure/case_studies_arxiv_v2.pdf
deleted file mode 100644
index 291c3a6..0000000
Binary files a/_site/src/figure/case_studies_arxiv_v2.pdf and /dev/null differ
diff --git a/_site/src/figure/case_studies_arxiv_v2.png b/_site/src/figure/case_studies_arxiv_v2.png
new file mode 100644
index 0000000..98893e5
Binary files /dev/null and b/_site/src/figure/case_studies_arxiv_v2.png differ
diff --git a/_site/src/figure/crafter_condensed_results.pdf b/_site/src/figure/crafter_condensed_results.pdf
deleted file mode 100644
index 7177512..0000000
Binary files a/_site/src/figure/crafter_condensed_results.pdf and /dev/null differ
diff --git a/_site/src/figure/crafter_condensed_results.png b/_site/src/figure/crafter_condensed_results.png
new file mode 100644
index 0000000..c17e445
Binary files /dev/null and b/_site/src/figure/crafter_condensed_results.png differ
diff --git a/_site/src/figure/fig1_iclr_e_notation.pdf b/_site/src/figure/fig1_iclr_e_notation.pdf
deleted file mode 100644
index 63fc3e8..0000000
Binary files a/_site/src/figure/fig1_iclr_e_notation.pdf and /dev/null differ
diff --git a/_site/src/figure/fig1_iclr_e_notation.png b/_site/src/figure/fig1_iclr_e_notation.png
new file mode 100644
index 0000000..86f80e9
Binary files /dev/null and b/_site/src/figure/fig1_iclr_e_notation.png differ
diff --git a/_site/src/figure/popgym_summary_expanded_outliers.pdf b/_site/src/figure/popgym_summary_expanded_outliers.pdf
deleted file mode 100644
index 909991a..0000000
Binary files a/_site/src/figure/popgym_summary_expanded_outliers.pdf and /dev/null differ
diff --git a/_site/src/figure/popgym_summary_expanded_outliers.png b/_site/src/figure/popgym_summary_expanded_outliers.png
new file mode 100644
index 0000000..7d8d5db
Binary files /dev/null and b/_site/src/figure/popgym_summary_expanded_outliers.png differ
diff --git a/_site/src/logos/rpl_logo.pdf b/_site/src/logos/rpl_logo.pdf
deleted file mode 100644
index ce3826d..0000000
Binary files a/_site/src/logos/rpl_logo.pdf and /dev/null differ
diff --git a/_site/src/logos/rpl_logo.png b/_site/src/logos/rpl_logo.png
new file mode 100644
index 0000000..89195ff
Binary files /dev/null and b/_site/src/logos/rpl_logo.png differ
diff --git a/index.markdown b/index.markdown
index 17b685e..cce3a08 100644
--- a/index.markdown
+++ b/index.markdown
@@ -158,7 +158,7 @@ highlight {
-AMAGO: Scalable In-Context Reinforcement Learning for Adaptive Agents
+AMAGO: Scalable In-Context Reinforcement Learning
for Adaptive Agents
Jake Grigsby1
Jim Fan2
@@ -204,7 +204,7 @@ AMAGO improves memory and adaptation by optimizing long-context Transformers on
@@ -228,7 +228,7 @@ In-Context RL's flexibility lets us evaluate AMAGO on many generalization, memor
-
+
@@ -243,7 +243,7 @@ AMAGO handles meta-learning as a simple extension of zero-shot generalization, a
-
+
@@ -270,7 +270,7 @@ As an example, we evaluate instruction-conditioned agents in the procedurally ge
-
+
@@ -329,7 +329,7 @@ Above, we use several single-task instructions to evaluate the exploration capab
In-context RL is applicable to any memory, generalization, or meta-learning problem, and we have designed AMAGO to be flexible enough to support all of these cases. Our code is fully open-source and includes examples of how to apply AMAGO to new domains. We hope our agent can serve as a strong baseline in the development of new benchmarks that require long-term memory and adaptation. Check it out on GitHub here.
-
+
diff --git a/src/figure/case_studies_arxiv_v2.pdf b/src/figure/case_studies_arxiv_v2.pdf
deleted file mode 100644
index 291c3a6..0000000
Binary files a/src/figure/case_studies_arxiv_v2.pdf and /dev/null differ
diff --git a/src/figure/case_studies_arxiv_v2.png b/src/figure/case_studies_arxiv_v2.png
new file mode 100644
index 0000000..98893e5
Binary files /dev/null and b/src/figure/case_studies_arxiv_v2.png differ
diff --git a/src/figure/crafter_condensed_results.pdf b/src/figure/crafter_condensed_results.pdf
deleted file mode 100644
index 7177512..0000000
Binary files a/src/figure/crafter_condensed_results.pdf and /dev/null differ
diff --git a/src/figure/crafter_condensed_results.png b/src/figure/crafter_condensed_results.png
new file mode 100644
index 0000000..c17e445
Binary files /dev/null and b/src/figure/crafter_condensed_results.png differ
diff --git a/src/figure/fig1_iclr_e_notation.pdf b/src/figure/fig1_iclr_e_notation.pdf
deleted file mode 100644
index 63fc3e8..0000000
Binary files a/src/figure/fig1_iclr_e_notation.pdf and /dev/null differ
diff --git a/src/figure/fig1_iclr_e_notation.png b/src/figure/fig1_iclr_e_notation.png
new file mode 100644
index 0000000..86f80e9
Binary files /dev/null and b/src/figure/fig1_iclr_e_notation.png differ
diff --git a/src/figure/popgym_summary_expanded_outliers.pdf b/src/figure/popgym_summary_expanded_outliers.pdf
deleted file mode 100644
index 909991a..0000000
Binary files a/src/figure/popgym_summary_expanded_outliers.pdf and /dev/null differ
diff --git a/src/figure/popgym_summary_expanded_outliers.png b/src/figure/popgym_summary_expanded_outliers.png
new file mode 100644
index 0000000..7d8d5db
Binary files /dev/null and b/src/figure/popgym_summary_expanded_outliers.png differ
diff --git a/src/logos/rpl_logo.pdf b/src/logos/rpl_logo.pdf
deleted file mode 100644
index ce3826d..0000000
Binary files a/src/logos/rpl_logo.pdf and /dev/null differ
diff --git a/src/logos/rpl_logo.png b/src/logos/rpl_logo.png
new file mode 100644
index 0000000..89195ff
Binary files /dev/null and b/src/logos/rpl_logo.png differ
In-context RL is applicable to any memory, generalization, or meta-learning problem, and we have designed AMAGO to be flexible enough to support all of these cases. Our code is fully open-source and includes examples of how to apply AMAGO to new domains. We hope our agent can serve as a strong baseline in the development of new benchmarks that require long-term memory and adaptation. Check it out on GitHub here.
- + diff --git a/_site/src/figure/case_studies_arxiv_v2.pdf b/_site/src/figure/case_studies_arxiv_v2.pdf deleted file mode 100644 index 291c3a6..0000000 Binary files a/_site/src/figure/case_studies_arxiv_v2.pdf and /dev/null differ diff --git a/_site/src/figure/case_studies_arxiv_v2.png b/_site/src/figure/case_studies_arxiv_v2.png new file mode 100644 index 0000000..98893e5 Binary files /dev/null and b/_site/src/figure/case_studies_arxiv_v2.png differ diff --git a/_site/src/figure/crafter_condensed_results.pdf b/_site/src/figure/crafter_condensed_results.pdf deleted file mode 100644 index 7177512..0000000 Binary files a/_site/src/figure/crafter_condensed_results.pdf and /dev/null differ diff --git a/_site/src/figure/crafter_condensed_results.png b/_site/src/figure/crafter_condensed_results.png new file mode 100644 index 0000000..c17e445 Binary files /dev/null and b/_site/src/figure/crafter_condensed_results.png differ diff --git a/_site/src/figure/fig1_iclr_e_notation.pdf b/_site/src/figure/fig1_iclr_e_notation.pdf deleted file mode 100644 index 63fc3e8..0000000 Binary files a/_site/src/figure/fig1_iclr_e_notation.pdf and /dev/null differ diff --git a/_site/src/figure/fig1_iclr_e_notation.png b/_site/src/figure/fig1_iclr_e_notation.png new file mode 100644 index 0000000..86f80e9 Binary files /dev/null and b/_site/src/figure/fig1_iclr_e_notation.png differ diff --git a/_site/src/figure/popgym_summary_expanded_outliers.pdf b/_site/src/figure/popgym_summary_expanded_outliers.pdf deleted file mode 100644 index 909991a..0000000 Binary files a/_site/src/figure/popgym_summary_expanded_outliers.pdf and /dev/null differ diff --git a/_site/src/figure/popgym_summary_expanded_outliers.png b/_site/src/figure/popgym_summary_expanded_outliers.png new file mode 100644 index 0000000..7d8d5db Binary files /dev/null and b/_site/src/figure/popgym_summary_expanded_outliers.png differ diff --git a/_site/src/logos/rpl_logo.pdf b/_site/src/logos/rpl_logo.pdf deleted file mode 100644 index ce3826d..0000000 Binary files a/_site/src/logos/rpl_logo.pdf and /dev/null differ diff --git a/_site/src/logos/rpl_logo.png b/_site/src/logos/rpl_logo.png new file mode 100644 index 0000000..89195ff Binary files /dev/null and b/_site/src/logos/rpl_logo.png differ diff --git a/index.markdown b/index.markdown index 17b685e..cce3a08 100644 --- a/index.markdown +++ b/index.markdown @@ -158,7 +158,7 @@ highlight {
-
+AMAGO: Scalable In-Context Reinforcement Learning
Jake Grigsby1
Jim Fan2
@@ -204,7 +204,7 @@ AMAGO improves memory and adaptation by optimizing long-context Transformers on
@@ -228,7 +228,7 @@ In-Context RL's flexibility lets us evaluate AMAGO on many generalization, memor