An agent that carries memory across tasks is only as good as the entries it keeps. We begin from an observation that is easy to reproduce and uncomfortable for the field: naively accumulating cross-task experience can make an agent worse than one with no memory at all. On AgentBoard, a workflow-induction memory solves fewer tasks than a memoryless ReAct agent in seven of eight domains, and a per-step procedural store with no surrounding structure loses half of ReAct's completions on Jericho. We argue that the missing ingredients are per-action credit assignment and an explicit life cycle for stored entries, and present AdMem, which supplies both. Before each action the actor commits to an expected outcome; a concurrently running critic scores the realized outcome against that expectation and writes a procedural entry carrying a binary reward and a reflection, so credit lands on the action that earned it and no environment reward is required. Each entry then carries a scalar utility, updated online under a noisy-OR reward model, that ranks retrieval jointly with context similarity and drives eviction and merging, giving the store a forgetting and consolidation policy instead of unbounded growth. Across 707 tasks in eight AgentBoard domains AdMem solves 441 against 388 for ReAct and 312 for the workflow-induction baseline --- a pooled gain significant under a two-sided Fisher exact test, concentrated in the domains that supply transferable cross-task structure. All memory stays human-readable text, which is what makes the failures above diagnosable --- and, as we discuss, also makes the store an auditable but attackable surface.
Figures & tables
Figure 1: Information flow between the environment, the actor, the critic, and the long-term memory store. The three modules run in separate threads: the actor never blocks on the critic, so memory writing and evaluation overlap with acting. The actor’s short-term memory (STM) holds the task instructions, the task and memory tool specifications, and the current task’s short-term trace, with environment facts surfaced through a priority-ranked semantic heap and event summaries through a scope-ordered episodic stack ( Section 2.3 ); retrieved long-term memory (LTM) enters the same context. Every action the actor sends to the critic carries the expectation the actor committed to before acting, which is what the critic scores the realized outcome against; the resulting per-action reward travels to the store alongside the new procedural entry and updates the utilities of the entries that were retrieved for that action ( Section 2.4 ).
Domain
AlfWorld
BabyAI
Jericho
PDDL
Number of tasks
134
112
20
60
Metric
C
P
C
P
C
P
C
P
ReAct
49.3%
0.6586
97.3%
0.9732
40.0%
0.4002
33.3%
0.5537
AWM
47.0%
0.5467
92.9%
0.9375
10.0%
0.2227
43.3%
0.6407
AdMem
63.4%
0.7755
100.0%
1.0000
45.0%
0.5569
76.7%
0.9069
Domain
Tool-query Academic
Tool-query Weather
WebShop
ScienceWorld
Table 1: Task completeness (C, the fraction of tasks solved) and average progress (P, the mean of the per-task progress score pi∈[0,1] ) on the eight AgentBoard domains, using Claude Haiku 4.5 as the backbone for every method. Bold marks the best value in each column. Each domain is a fixed task set of the size given in the second row, so completion counts are recoverable exactly from C (e.g. 63.4% of 134 is 85 tasks); we use those counts for the Fisher exact tests reported in the text. Each configuration is streamed once over the task set in a single lifelong episode. On the three 20-task domains a single task is worth 5.0 points of C, which is why we base per-domain conclusions on the tests rather than on the point differences alone.
Method
Tasks solved / 707
95% CI on C
Macro-avg. C
Domains ≥ ReAct
ReAct
388 (54.9%)
[51.2, 58.5]
56.4%
—
AWM
312 (44.1%)
[40.5, 47.8]
41.7%
1 of 8
AdMem
441 ( 62.4% )
[58.7, 65.9]
69.2%
6 of 8
Table 2: Aggregate view of Table 1 over all 707 tasks. Intervals are Wilson 95% binomial intervals on the pooled completion rate. Macro-avg. C weights the eight domains equally rather than by task count; we report both because WebShop and ScienceWorld together hold 48% of the tasks. Against ReAct, the pooled difference is significant for AdMem ( +53 tasks, two-sided Fisher exact p=5×10−3 ) and significant in the negative direction for AWM ( −76 tasks, p=7×10−5 ). Because all three methods are run over the same 707 tasks, both comparisons are paired; both remain significant under the least favorable exact McNemar test consistent with these margins ( p=0.03 and p=5×10−3 respectively, Appendix C ).
Domain
Jericho First epoch
Jericho Second epoch
Jericho Third epoch
Metric
C
P
C
P
C
P
ReAct
40.0%
0.4002
-
-
-
-
STP
40.0%
0.4553
-
-
-
-
LTP
20.0%
0.4257
25.0%
0.4775
30.0%
0.4743
AdMem
45.0%
0.5569
50.0%
0.6578
60.0%
0.6820
Table 3: Component ablation on Jericho (20 tasks), Claude Haiku 4.5. STP : short-term planning and context management only, with all long-term recall removed. LTP : the long-term procedural store only — per-step retrieval with critic-generated reflections and the reward-based utility retained — with short-term planning and context management removed. AdMem : all components. STP and ReAct are single-epoch by construction, since neither carries anything across tasks. For LTP and AdMem the task set is streamed again each epoch without resetting the long-term store, so the epoch axis measures accumulation rather than repetition of a fixed store; entry utilities carry over with the store, so an entry that proved unhelpful in an earlier epoch enters the next one already down-weighted.