Mod: Modded Radious alone.
I need a specific set of issues to be looking for.
What is the intent in changing each table?
All I can say is something is off compared to my last play throughs with my personal mod of Radious. I can't detail more until I know whether to assess blame to chance or a change in the this mod.

this mod was never used or tested on its own like you have using it. This has been explained in the description. If you read the description you would know "It has been developed, tested and balanced with "Radious - Very Hard Campaign Overhaul" (It's basically a submod"

if you see any bugs please provide more details and I will try to repair it but as for now all I see is "something is off". This will not help in finding errors

I started a WRE campaign on hard using your mod and Radious with larger garrisons plus Radious total units.

I renumbered your tables names to overwrite the radious/my own tables. How did I do that?
8 turns in here's what SEEMS to be different but could be just random:
The AI is aggressive, but not crazily so.
In normal mode, they usually find a neighbor plus Rome to attack. In this mod, they all picked Rome until around turn 5 when they found they could attack vandals.
The AI attacks in dual armies. In normal mode, they usually leave at least a mini stack at home.
With this mod, they leave their base unguarded.
At turn 8, all the AI seems to have build level 2 town, 1 eco, 1 religion. In normal, they usually bee line towards chieftain houses and barracks. This is good, but it leads to large groups of trash units. I think it would be better if they all didn't build religious buildings.
Your mod seems the opposite of XAI in some ways. Their the AI guards their base closely and has trouble deciding whether to attack with 1 stack.
Also, the rebel armies are waiting til they reach 20 units. I used to count on a battle at about 15 units.
All armies seem to seek overwhelming numeric advantage before attacking.

AI seems to place little weight on relationship level before declaring war. I was friends with 2 factions (green or plus 35) but was attacked without even an alliance war. The AI does this to each other. After I defeated a German faction double stack, reducing their strength to 0, their neighbor, a green friend, attacked them and took their home base

The AI is smart in playing against humans. This is my worst start ever as WRE/hard. But I wonder if it is realistic for them not to guard their home towns. or spend more on construction. It seems they are spending about 80% of funds on junk stacks, attacking then building more junk stacks.

Other notes: The saxons have razed twice. There are no agents that I can see from the AI.

So is any of this your mod? Which tables?

2 turns later and I've spied around much farther. The barbarian buildings are actually quite diverse and seem to be fairly balanced. The barbs are creating agents at a slow and reasonable pace versus Vanilla. The hordes are having trouble settling down for more than one turn and rebuilding. They are not under attack. This might be a flaw implicit in all Radious and Vanilla games.

yes it seems like all those results coming from my AI mod. Now for Radious this mod is just a submod of overhaul so I never tested it on like you had. I only play "Radious - Very Hard Campaign Overhaul" But I can see this aggressiveness of AI is definitely coming from this mod. The tables that makes AI more aggressive and calculates how much of army should take on you can find in "cai_personality_strategic_components" and "cai_variables". It starts in table 10. If you play with those ratios the results in tests plays will show you response of AI making different choices. I have decided to keep my AI agressive rather than stay at home and guard all the time (Its not like my AI doesnt do that anymore , because it does a lot but not as much) because if AI does than it never expands and creates litlle empires wich makes it very weak compared to a player. Late game is so boring then and it makes no sense and it gives me no fun . That is why we have strong garrisons mods to keep home safe and occasionally 1 stack on top.

1. What garrisons mod are you using for Radious if I may ask ?
2. And have you tried my overhaul "Radious - Very Hard Camapaign Overhaul" with "Radious - Garrison [unofficial]" ?
3. If not then Have you tried Radious together with my submods like "AI"+"Diplomacy"+"Radious - Difficulty"+"Progress"+""Autoresolve"+"Movement"+"Empire" ? - because those mods were designed at the exactly same time as I was tweaking the AI mod so your results might look more satisfying when actually mods are not just about AI but the whole game system.

ps. I know what you thinking that you could maybe tweak it to make AI so smart that it will guard their home and attack when it has to. But no , trust me AI in this game is ♥♥♥♥! game engine would never allow you to actually make AI smart. If you make it defensive it will get weak , if you make it aggressive it will get strong by chance but leave its home unguarded more often than not. Compromise is not possible in this game I've tried. If you say that others have made it like Xai (not saying I have tried his mods thoroughly yet)for example , can you actually say that you have made it into late game like 450 year or so ? If yes then what was the outcome? Player always wins ? that's kind of not my style of playing games. ; I like to win when there is not much of a chance of winning the game... playing on "Very Hard" and optimising my mods on that difficulty. Hard and Normal is for people who never played total war games actually or have very little experience.

https://steamuserimages-a.akamaihd.net/ugc/955210546398828793/E809A750BDEF19FFE901A9C6A88C2ADC132302BC/

^ this picture has been taken now on my playtest with This AI mod! and other mods that I have mentioned and you can find it all in my 1click collection vanilla. I have played without Radious and once again AI shows that actually when it has to it will defend its own territory. On this picture we have already year 400 and I play Geats on Very Hard. 5 factions around me are not at war with me but they are guarding territories even though they are at war with other factions. They even come to border sometimes patrolling/consolidating. I am not going to test my mod on its own as a standalone mod because that would simply not be a good idea and results would be completely different I'm sure of that.

I did a comparison of the AI tables.
You indeed give extra bonuses for defending faction capital.
I disagree with some of your values based on my experience with these tables and research on the forums. But overall, it looks pretty good.
My question: I keep hinting at your naming convention. Why don't you fix it? How do you know that you are even testing your mod when combined with others?
=====================================Table Analysis for AI========================================
cai_personalities_budget_allocations
1:Key
2: What can spend on agents per turn. for larger factions
3: Cap based on percent of total funds. So can't recruit 1000gold spy unless are making 10000 if it is set at 10%. Restricts to big factions effectively.
I wonder if there is a pool that builds over 4 turns. I think it should at least equal the cost of an agent after the pool timer elapses.
4: What is on fail?
5: Pool builds over 4 turns? If so, this allows us to have agents that only take actions once pool fills up sufficiently. Expand the pool lower the allocation
6" For big factions. This is how the Sassanids throw 5 stacks at you as ERE. I would lower it somewhat since the warweariness fx is broken
7 All factions in last stand mode should have high military spending. But what sets the last stand?
Note that if increase here, need to decrease allocation% by same amount elsewhere.
8 Percent of pool to save??
9 Turns of Inactivity until cap
10 11 Construction fund allocation% Tempted to raise this but what about the small factions who have nothing to spendon.
Diplomacy funding cap: Data has this at 1000. Radious at 1000. I think it should be around 5-10K to see negotation take place. If set too high,
the AI might offer too much money and get ripped off.
Allocation percentage is much more important here. If set so low, I would keep it at 10K

=====================================================
cai_personalities_income_allocations
I am guessing that when this is too high the ai won't have money to build or convert larger buildings.
This is effectively a military/diplomacy/agent table. Lower the proportion to get buildings?
I think Vanilla has it right and would not tamper with raising it.
====================================================
cai_personality_deal_generation_generator_priorities
diplomat DECLARE_WAR_ON_FACTIONS_I_DISLIKE 20 0 -190 0 0 190 190 110 140 5 0 10000 35
Lower values in Priority columns indicate less likely to act on this priority
Better to have them declare war locally then chase enemy across the map.
Better to have higher trade.
Reset DECLARE_WAR_ON_NEARBY_FACTIONS to Vanilla
Lowered DECLARE_WAR_ON_FACTIONS_I_DISLIKE by about 20-30 each so could increase declare war on nearby factions a larger degree.
Trade Agreements: Increase across the board
OFFER_TO_JOIN_ALLYS_WAR I don't see the point when can just increase REQUEST_ALLY_TO_JOIN_WAR
=====================================================

cai_personality_negotiation_components
Does not seem worth my time. Not sure it is implemented well in game.
=====================================================
cai_personality_occupation_decision_priorities
I have tested this one for 2 years.
I set razing at 1 for burn it all and still see some factions raze that shouldn't.
Resettling: doesn't that need to be higher than 1? Some mods set it at 200.
The problem with that is that you get factions settling all over the map.
Colonise: Again, some mods put this at 200 for all. I believe it works fairly well in the 15-50 range.
occupation_decision_gift_to_another_faction: I've never seen it.
occupation_decision_liberate: this happens far too much, so I've lowered it dramatically.
occupation_decision_occupy: We need to see more of this for certain factions ex the hordes and distant factions.
Vassal: This should be at minimal levels. An AI vassal becomes like a 2nd owned faction I have yet to see it rebel.
======================================================
cai_personality_strategic_components
With so many unknown columns, I am afraid that I'll screw up part of a formula. Don't touch from Vanilla.
======================================================
cai_personality_strategic_desired_attitudes

Vanilla values are set at very precise levels indicating that someone put alot of work in it.
I don't agree that one should decrease the bonuses of friendships and alliance vis a vis pos and neg factors.
Keep Vanilla
----------------------------------------------------------
==================This table seems important======================================
cai_task_management_system_task_generator_groups_generators_junctions
The "Actively Defend" fields are given very high priorities. I don't see this happening in-game.
Could it be that the priority numbers are out of bounds. Vanilla uses values between .01 and 1.5 typically.
https://steamhost.cn/steamcommunity_com/workshop/filedetails/discussion/193204000/1697169163405151788/
CA actually found better results with smaller boundaries. It seems they are still working on it given the recent developer post. Maybe look at Rome2? Warhammer?

CAI_TMS_TASK_GENERATOR_ATTACK_ALL_ENEMY_FORCES I would be careful with this one. There are many other tables that set up aggressiveness with reasons. Raised .15 each
Hordes need to be careful because they face wipeout.

CAI_TMS_TASK_GENERATOR_ACTIVELY_DEFEND_CAPITAL is important due investment, but it is hard to gain any victories if not double teaming a settlement at least for 1-2 turns.

CAI_TMS_TASK_GENERATOR_ATTACK_ALL_FLAGGED_VICTORY_CONDITION_REGIONS Does the AI really care? Player is the only one that can have victory. Use Vanilla.
CAI_TMS_TASK_GENERATOR_ATTACK_LOCAL_FLAGGED_VICTORY_CONDITION_REGIONS

CAI_TMS_TASK_GENERATOR_ATTACK_ALL_LOCAL_HUMAN_SETTLEMENTS I wouldn't increase this more than already is. Increase the Human forces
CAI_TMS_TASK_GENERATOR_ATTACK_FOOD_PRODUCING_REGIONS Aren't there more important priorities? AI gets a big food bonus already.
CAI_TMS_TASK_GENERATOR_ATTACK_HORDE_ARMIES_HUMAN_AT_WAR_NOT_SETTLED It's hard enough just to keep a horde alive as a human Vanilla
Theme: Make it easier to play as a horde. They have enough problems just rebuilding armies.

CAI_TMS_TASK_GENERATOR_ATTACK_NEIGHBOURING_WAR_REGIONS Use Radious
CAI_TMS_TASK_GENERATOR_ATTACK_RECENTLY_SACKED_SETTLEMENTS It is a problem that the game lets this happen. 1 Sack should wipe out a settlement. Vanilla
CAI_TMS_TASK_GENERATOR_DEFEND_ALLIED_TERRITORY This happens too much. Vanilla a bit too weak. Use Radious values of 1
CAI_TMS_TASK_GENERATOR_HORDE_FOLLOW_LEAD_ARMY Increase somewhata, but this decreases survival chance if wander into bad region.
CAI_TMS_TASK_GENERATOR_RAID_UNSTABLE_REGIONS Historically, this would yield little loot. use Radious

CAI_TMS_TASK_GENERATOR_SACK_SETTLEMENT_TO_IMPROVE_MORALE Improve morale more by raiding

CAI_TMS_TASK_GENERATOR_WAR_COORDINATION_ATTACK_FORCES Increase from Vanilla as this is often ignored.
CAI_TMS_TASK_GENERATOR_WITHDRAW_DAMAGED_ARMIES As Rome, too many damaged armies wander around and raid. Why don't they go home and replenish? Radious
CAI_TMS_TASK_GENERATOR_WITHDRAW_LOW_MORALE_ARMIES Decrease this from Vanilla as it's better to wipeout and start over. Also, small factions have imperium penalties that lower their morale even when they have greater numbers than enemy
==============================================================================
CAI_Variables:
CAI_VARIABLE_AGGR_RAMP_ENABLED 0 (IN GAME) Whether the aggression ramp is enable. 0=no, 1=yes
Vanilla turned it off for good reason, so why bother with it?
Minos and Greq did good job here choosing values, but fear some may be out of bounds.

Thank You for your analysis. I will slowly answer all your points.
My naming conventions of what table you mean ? or you mean the name of a mod ? It loads before anything else I use so I'm not sure what you meant by that.

4. I think it's the actual "fund" for seperate stuff like agents in this case. Fail will happen for number of reasons. One of which can be change of priorities for example. In other table cai_personalities_income_allocations you can see percentage of whole income going into different pools. If it reaches enough for recruiting one agent in 4 turns faction will recruit. If not it will save 75% for next try.
5. Why should I do this? I don't see problems with agents in my game. AI uses its agents so its all good I think
6. I like to see rich AI like sassanids sends few stacks of army at my settlements. If Ai would not do that then player always wins because he can do it so why not AI ? I don't get this idea of making AI poor in armies. AI should be aggressive and it should conquer settlements in my opinion.
7. Last stand is when faction lost their all settlements down to last one or horde all armies down to last one. and if you look closely I have intentionally gave more % for militarists factions so they can recruit quickly more armies to fight back. but other factions will spend more evenly because they can use diplomacy to ask for peace for example instead fighting to the last blood.
8. same like with agents. AI in this mod will save more for army at last stand even if fails to recruit for example military buildings were destroyed ect.

Diplomacy funding cap is 20000 because I found it's the most optimal and Ai in my mods never pays too much trust me. This for situations when you have lots more power than AI and AI respect you a lot + it has lots of money then it is possible it will pay a lot more in deals

cai_personality_deal_generation_generator_priorities

I have slightly moved towards Vanilla with declaring wars. You are right. This table was overlooked . Can you tell me what actully "Total_War"_priority means exactly ? Is this when it AI" has wars with many factions including player at once?
OFFER_TO_JOIN_ALLYS_WAR I find this more interesting during a campaign when AI actually wants to help you sometimes for a bit of money ofc. It works just fine because I've seen it and AI is not abusing it.

cai_personality_negotiation_components
it is implemented . Diplomacy is different than in vanilla

cai_task_management_system_task_generator_groups_generators_junctions

When I have time Ill test it with smaller values but from my previous tests AI is just very passive and not willing to attack or do much if the values are small. But as I said I cannot be 100% sure because small tweak here and there and it all changes. From what you are saying my Ai would be very confused but it isnt. You might be right about the hordes though they are dying very quickly but in vanila they also do and other AI mods did not improved on that.

Good to see you thinking. @_ symbol has lower priority then numbers. Can you verify that it has lower priority than letters? I use the fame table across several mods to compare. You can see next turn up top what the result it.

@_ loads before all mods I use basically. So it seems good enough. Numbers can be used for patches and such. Also it's too late to change the name even if I wanted to.

Since we both spend so much time on the same issue, why don't we collaborate and discuss this privately. I've been working on my own AI mod that I think works better than this. But I can hand it over to you.
I still am testing you: Which loads first @_ or 00? For clarification: I am referring to the mod table with the highest precedence.